AIRA

Page 1

1

INTRODUCTION:

Artificial Intelligence (AI) is currently leading a technological revolution that has significant implications for diverse sectors of society, ranging from high school students to CEOs. As an integral component of human innovation, AI has emerged as a transformative technology that has disrupted and reshaped traditional approaches to creating anddeliveringproducts.However, the rapid global expansion of AI companies has given rise to critical concerns regarding their handling of sensitive information, particularly in relation to issues of privacy and regulation. In response to these and other concerns, the Artificial Intelligence Regulation Authority(AIRA),a nonprofit organization, has been established to provide a platform for regular global citizens to engage in constructive deliberations on the regulations governing AI companies. The central objective of AIRA is to ensure that power is distributed equitably, enabling all individuals who may be affected by AI to participate in decision-makingprocessesthataddresscriticalquestions relatedtotheregulationofAIcompanies.

WHATWESTANDFOR:

1. The transparency of Artificial Intelligence technology and usage: We want citizens to have a voice in how AI will affect their lives, and we want AI companies to be transparent with the public in all of their decisions and actions. Transparency engenders trust, promotes accountability, and allows users to make informed decisions. Additionally, it can facilitate effective regulatory and compliance mechanisms, enabling the ethical and legal standards for AI usage to be upheld.Byupholdingtransparency,AI companies can demonstrate their commitment to responsible AI usage and establish meaningfulpartnershipswiththebroadercommunity.

2. The accountability of Artificial Intelligence companies: We hold the view that AI companies must be held responsible for any actions deemed to be in violation of the ethical codes and regulations that have been established by the citizens of the AIRA assemblyafterdeliberation.

3. The participation of ordinary citizens in determining the scope of Artificial Intelligence's capabilities and limitations to meet their needs: We hold the perspective that the primary stakeholders of Artificial Intelligence are ordinary citizens, who are either currently using, are in the process of becoming AI users, or are being affected by AI in any capacity. Therefore, we assert that these individuals should be the ones to determinetheregulationsthatoughttobeimposeduponAIcompanies.

4. Trust: We recognize thatthedigitalsocietymustbebuiltontrustamongallstakeholders, including governments, civil society, international organizations, academics, and =businesses, by sharing common values and principles, including equality, justice,

2

transparency, and accountability, while taking into account the global economy and interoperability.

5. Futuristic mindset: We believe in recognizing the possibility of future spheres of knowledgethatareyetunknown.

6. A never-ending conversation: We believe in and support a continued conversation and diversity of views on how to achieve an inclusive, sustainable, safe, trustworthy, and innovativesocietythroughdigitalization.

7. Artificial Intelligence for the safety of all: We believe in the use of AI technologies to promote inclusive economic growth, bring great benefits to society, and empower individualswithresponsibledevelopmenttomitigateriskstowidersocietalvalues.

8. Benefits of Artificial Intelligence: We have a strong recognition of the benefits brought by the responsible use of AI to improve the work environment and quality of life, and create potential for realizing a human-centered future society with opportunities for everyone,includingwomen,girls,andvulnerablegroups.

9. The importance of government: Werecognizetheimportanceforgovernmentsandother stakeholders within their respective roles to addresssecuritygapsandvulnerabilitiesthat have a negative impactondigitalinnovationsandtrustbyconsumersandbusinesses,and thus hinder us from taking full advantage of the benefitsofdigitalization.Securityinthe digitaleconomyisalsoimportantforgovernmentsinprovidingtheirservices.

10. Connectivity and broadband access: We recognize that improved connectivity and broadband accessarenecessaryconditionsforthedevelopmentofthedigitaleconomy,as wellaspowerfulenablersofinclusivegrowthandsustainabledevelopment.

11. Policy approaches: Weacknowledgetherelevanceofappropriatepolicyapproachesthat, while recognizing national circumstances, promote a pro-investment, fair, competitive, and non-discriminatory marketplace, enhance the accessibility, affordability, quality, and security of connectivity and digital services, and increase access to digital economic growth.

12. Digitization: We acknowledge the role and ability of digitization to contribute to economicgrowthandsocialdevelopmentinavarietyofsectors.

13. Cooperation between ordinary citizens, executives, and experts: Our perspective advocatesforacollaborativedecision-makingprocessthatinvolvesboththeend-userand theproducertoensuremutualcooperationandco-creation.

14. The power of collective intelligence: Westronglybelievethatthecollectiveinvolvement of individuals from diverse backgrounds, withequalparticipationinthedecision-making process, leads to better decision making due to cognitive diversity and avoidance of biases, and leads to outcomes that are more conducive to achieving organizational and humanobjectives.

OVERVIEWOFEACHSECTION:

3

This document will present our vision for AIRA in three parts. The first part covers our motivation as well as the current background and theoretical context AIRA operates in. Section 1.1 lays out what AI is and why AI governance is important. Section 1.2 paints the existing regulatory backdrop for AI, which AIRA will fit into. Section 1.3 elucidates the key argumentative strands for why sortition is the main governance principle behind AIRA, as well as the current examples of sortitionusedforgovernance,whichAIRAisinspiredby.Section1.4 details jurisdiction, where the proposal covers who AIRA will regulate–in short, AI companies–aswellastheethicalstandardsandcodesthatAIRAwillholditselfto.

The second part presents our proposal for AIRA. It begins with Section 2.1 going into the functions of AIRA’s three branches: the General Assembly (Legislative), the Governance Committee (Executive), and the International Arbitration mechanism (Judiciary). The second partclosesoffwithSection2.2whereweanticipatepotentialchallengestoourproposedmodel.

Finally, in the third part, we end with Section 3.1 which discusses the implementation of AIRA–how it interacts with and aligns with the incentives of AI companies, existing AI regulationsandotherstakeholders.

Section1.1: What is AI

● WhatisthedefinitionofArtificialIntelligence(AI)?

● WhatisthesignificanceofAIgovernance,andwhyisitconsideredacrucialelement?

Section1.2:ExistingLegalFramework

● WhatarethecurrentframeworksavailableforAIgovernance?

● InwhatwaysdotheseframeworksfallshortintheirabilitytoeffectivelygovernAI?

Section1.3: Why Sortition

● What are the reasons that support the superiority of deliberative democracy as a governingframeworkforAI?

● What factors will contribute to establishing the legitimacy of the AI Regulatory Agency (AIRA)?

Section1.4: Jurisdiction

● WhichentitiesorindividualsaresubjecttoregulationbyAIRA?

● WhatisthemechanismforestablishingtheagendaunderthepurviewofAIRA?

4

Section2.1: AIRA Structure

● Whichindividualsorgroupsareencompassedinthescopeofinclusion?

● What is the selection criterion for the participants, for instance, will it be based on stratification?

● Whatisthedurationoftheconventionandthelengthofeachsession?

● Willtheconventionbeconductedvirtuallyorinperson?

● Whatisthecontributionofsubjectmatterexpertsintheconvention?

● What level of transparency will be upheld with the general public during theconvention sessions?

● What strategies will be employed to guarantee adherence to the legislation that hasbeen approvedbytheassembly?

● Whatmeasureswillbetakenifanybreachorcontraventionofthelegislationoccurs?

Section2.2: Potential Challenges

● Whatpotentialobjectionsareexpectedtoarise?

● Whatmeasureswillbeemployedtocounterorresolvetheidentifiedobjections?

Section3.1: Implementation

● Whatistheenvisagedapproachtoachievetheimplementationofthisproject?

● What strategies will be employed to secure the engagement and commitment of companiestotheproject?

● What incentives will be offered to AI companies that differentiate us from their current investmentinAIethicsandgovernance?

● What collaborative measures will be taken with the current AI regulatory organizations, such as Partnership on AI, and how will the existing AI legal frameworks be integrated intotheproject?

OURGOALSFORAIRA:

Our objective is to establish a forum that provides an opportunity for individuals from various regions of the world to convene and engage in discussions concerning the use of AI globally. Our aim is to foster an environment where citizens candeliberateonthemostpressing issues regarding AI and develop a collective solution to regulate AI companies and their capabilitieswhileprioritizingprivacyandhumanrights.

5

OVERVIEWOFTHEASSEMBLY:

Figure1:Overviewoftheassembly

GLOSSARY:

AI: Artificial intelligence (AI)isasetoftechnologiesthatenablecomputerstoperformavariety of advanced functions, including the ability to see, understand and translate spoken and written language,analyzedata,makerecommendations,andmore.

Sortition: Theactionofselectingordeterminingsomethingbythecastingordrawingoflots.

Accountability: The state of being responsible or answerable for a system, its behavior, and its potentialimpacts

Privacy: Thestateorconditionofbeingfreefrombeingobservedordisturbedbyotherpeople.

6

Cognitive Diversity: Cognitivediversityisacompany'svarietyofemployees—allofwhomhave differentperspectives,backgrounds,ethnicities,andeducationlevels.

PART1

SECTION1.1:WhyArtificialIntelligence

Artificial Intelligence, AI, is a form of technology thatmimicshumanintelligencethusenabling computers to learn from experience through systematic repetition and algorithmic training. The idea of AI was first presented by John McCarthy and Marvin Minsky during the Dartmouth Summer Research Project on Artificial Intelligence in 1956 (Anyoha, 2017). Shortly after, Edward Feigenbaum created a computerprogram,EPAM:ElementaryPerceiverandMemorizer, which served as some of thefirstproofthatcomputerscouldlearn(Nilsson,2019).Inthe1980s, AI’s abilities continued to grow as a result of John Hopfield’s and David Rumelhart’s advancements in a technique known as “deep learning.” Their work essentially allowed computers to learn through experience, a skill previously only found in intelligentlife(Anyoha, 2017).

Nowadays, AI technology is “capable of significantly outperforming humans” and not just in a mechanical sense (“Why AI,” 2021). In the 1990s, Kismet, a robot head created at the Massachusetts Institute of Technology, became the first social robot, capable of “enter[ing] into natural and intuitive social interaction with a human” (“Kismet Overview,” n.d.). AI’s rapid evolution is indicative of its growingimpactontoday’ssociety.Presently,AIiscapableofdoing a repetitive task without flaw or fatigue, coming up with decisions, analyzing data, and maximizing the value of data at a higher efficiency and accuracy than a human and much cheapertoo(“WhyAI,”2021).

Such technology is currently employed in numerous fields such as healthcare, retail, manufacturing, and banking —inevitablytouchingmultipleaspectsofanindividual’slife(“Why AI,” 2021). In fact, as of2023,over47%ofestablishedorganizationshaveadefinedAIstrategy for mobile apps, 41% of consumers believe AI will benefit their lives, and 84% of global businesses believe that AI will give them a competitive edge (G., 2023). AIisnotonlygrowing smarter but it is also becoming woven into the fabric of society, raising the question ofwhento stop.

With the end goalofbeingabletoimitatehumanintelligenceandcapabilities,AIwilleventually start replacing human labor. In a study based on a dataset of 33 OECD countries it was found that both “robots and AI tend to increase unemployment” (Bordot, 2022). Thisinturnraisesthe question of whether or not there exists a future in which AI will diminish the value of ahuman

7

worker until it is essentially meaningless. In another study conducted by Daron Acemoglu, a professor in the Department of Economics at Massachusetts Institute of Technology, “AI and extensive use of data [is] likely to multiply automation possibilities,andthuscanexacerbatethe inequality trends the US and other advanced economies have experienced over the last several decades” (Acemoğlu, 2021). AI will also significantly increase competition in the economic field, widening the gap between “technological leaders and laggards in every sector. 'Early adopters', that is, companies that fully absorb AI tools over the next five to seven years, will mostprobablybenefitdisproportionately”(EuropeanParliamentaryResearchService,2019).

There are also various security concerns. For instance, when used asatoolinhealthcare,AIhas to have access to a wealth ofpatientinformationandissusceptibletoabreachinsecurity.These are just to name a few of the potential consequences of this ever-growing, and currently unbridled,field.

In a statement directly from OpenAI, one of the leading organizations in the field, the creators beg for action, understanding that suchtechnologymustberegulated(“PlanningforAGI,”n.d.). In an open letter, which was signed by well-known AI researchers as well as big names in the tech industry such as Elon Musk,asix-monthmoratoriumwasproposedforallAIlabscurrently “training AI systems more powerful than GPT-4 (“Pause giant AIexperiments,”2023).Theygo on to say that “this pause should be public and verifiable and include all keyactors”andcanbe used as time to “develop and implement a set of shared safety protocols” (“Pause giant AI experiments,” 2023). Open AI specifically requests that the institutions of the world agree on howtolimitthe“extremelywideboundsofhowAIcanbeused”(“PlanningforAGI,”n.d.).

AI is not an entity limited to a person or a place, but rather something that affects the entire world, consequently requiring global rather than simply national input on the parameters of its expansion.

SECTION1.2:ExistingLegalFramework

What policies governing AI currently exist at the national level?

Currently, very few laws existnationallyandinternationallytogovernArtificialIntelligence.No countries have specific laws governing the use of A.I. For example, the United States has no dedicated piece of legislation governing AI. Proposals have been put forward, such as the Algorithmic Accountability Act of 2022, but nothing has passed through the legislature (Algorithmic Accountability Act, 2022. In other countries, AI legislation has proceeded on an issue-by-issue basis instead of proactive legislationthatmovesinconjunctionwithAIevolution. For example, China banned AI-generated media without a watermark in December 2022 (Technica, 2022). Because of a lack of any concrete AI regulation, it issimplestforcountriesto

8

legislate on an issue-by-issue basis,reactingretroactivelytotheissuesalreadycreatedbyAI.No all-encompassing,broadlegislationhasbeenpassedbyanynation.

Due to the lack of expertise among legislators and the rapid evolution of AI, it is difficult for countries’ governments to legislate at the same speed that AI develops. Separately from governmental oversight, the U.K. has asserted that it will not create anationalAIpolicy;rather, it will rely upon independent companies to create tailored approaches to AI unique totheirown companies’ principles. This decentralized approach, characterized by the U.K. government as having the ability to tailor-make solutions to hyper-specific situations, creates the potential for exploitation (Browne, 2023). It leaves the decision to the whim of companies themselves. This presents a clear conflict of interest, as companies could potentially design AItoservemalicious ends that are not in the interestofthepublicsuchasviolatingprivacylaws.However,theU.K.’s approach points to the need for an expansive cohesive oversight of AI that can respond in an informed, timely, uniform manner If AI development and usage were regulated at their source by a global body focused on AI, instead of inconsistent attempts to rein in derivatives of AI usage from each country, the scrambletofindAIgovernancesolutionscouldbemitigatedforall countries.

There is little literature publicly available on AI regulation outside of the Global North, either because it is not publicized or because itdoesnotexist.Manycountrieswithlessinternetaccess and affordability than the rich countries in the Global North likely have no AI regulation. This regulatory hole leaves them vulnerable to exploitation, and leaves large proportions of the world’s population out of the global progression on AI. Moreover, as much of AI relies on crowdsourcing materials, underserved populations’ lack of widespread, easy access to the internet and AI will likely exacerbate the digital divide across the world. To shrink the digital divide and grow AI in an equitable manner, a global regulatory body that promotes and facilitatestheinclusionoftheopinionsofallcommunitiesaroundtheworldisnecessary.

What policies governing AI currently exist at the international level?

International governmental organizations also recognize the importance of a worldwide unified AI development policy, though nothing concrete has materialized. The European Union recognizes the General Data Protection Regulation (GDPR) as the de facto AI legislation currently, as it has no new legislation to supplant this piece. The GDPR protects the security of individuals’ personal data, and contains no specific reference to AI (EUGDPR,2022).Italyhas banned forms of AI, such as ChatGPT, on the grounds that it violated GDPR (Browne, 2023). The lack of legislation concerning AI has already impacted its global market and economic viability. At the end of 2022, the European Union adopted its general approach ontheArtificial Intelligence Act. Ifthelegislationsurvivestheprocesstobecomecodifiedintolaw,itwilltakeat least three years to be finalized (Drake & Ong,2023).TheEU’sproposedapproachdenotesthat the regulation would have extraterritorial reach, meaning that any AI system providing output

9

within the EU would be subject to the EU’s jurisdiction, regardless of where the AI provider company is located. This highlights the need for international cooperation on AI regulation and development, as the internet and AI transcend national boundaries. Looking to a broader global scope, member countries of the United Nations haveagreedtosignontoAIethicsstatutesinthe past. The UN’s Office of the Secretary-General’s Envoy of Technology declares that the way forward for members of the UN is, among other things, to create a “multi-stakeholder advisory body on global AI cooperation” (Providing Global Steerage, 2023). IGOs broadly acknowledge the need for international cooperative policy surrounding AI. A regulatory and advisory body with a global scope is necessary to further the advancement and beneficial use of AI equitably andethically.

What kinds of non-legal AI guidelines exist?

Many types of non-binding AI development guidelines have been developed both by governmental bodies and think tank research centers throughout the world, reflecting the need for coherent AI policy The U.S. Department of Commerce has released the Artificial Intelligence Risk Management Framework, anon-bindingdocumentintendedtoserveasaguide for safely and ethically using AI (NIST, 2023). Many other countries also have developed non-binding guidelines on AI, such as the Montréal Declaration foraResponsibleDevelopment of Artificial Intelligence in Canada and the Beijing Artificial Intelligence Principles (Montréal Declaration, 2018) (Beijing Artificial Intelligence Principles, 2022). The Montréal Declaration asks for signatories to join the pledge for responsible use and development of artificial intelligence (Montréal Declaration, 2018). All of these guidelines developed internationally recognizetheneedforaunifiedmannerofconductconcerningAIregulation.

What kinds of internal regulations do AI companies have?

Though some companies have experimented with democratic methods in regulation, most AI companies are not clear with their internal regulation on ethicsordevelopment,iftheypresenta strategy to the public at all. Using opacity as a deliberate corporate maneuver, many companies cloud knowledge on their self-imposed rules toallowforuninhibiteddevelopment,regardlessof theprivacyorotherrightsviolationsthedevelopmentofAImayincur(Burrell,2016).

Some companies have tried to create transparent and participatoryregulationmechanisms.Meta used Community Forums in 2022 to “make decisions that govern [their] technologies,” deliberately bringing together diverse groups of peoples to consider allperspectiveswithintheir technological development (Harris, 2022). These Community Forums utilized online deliberation, highlighting that direct participation and deliberationarekeytodigitalgovernance. In the past Meta has also used Oversight Boards, a system similartojudicialreviewintheU.S., to investigate whether content should be moderated or allowed within their systems (Levy, 2022). Within Meta, Facebook especially pushed for this system to introducehumanbeingsand users into the equation of regulating the systems they use. However, only Meta subsidiaries are

10

held to these standards since the regulations do not apply across the spectrum of AIcompanies. There is a need forregulationthatappliesequallytotheindustryasawhole.MostAIcompanies donotpresentastrategytothepublicforethicaldevelopmentanddiffusionofAIintotheworld.

SECTION1.3:WhySortition

AIRA’s foundationisbasedontheprinciplethatdemocracyisaforumnotamarket.Theconcept of a deliberative democracy is based on the ideaofademocraticcommunitywherecitizenswho are equal have discussions and use reasoning to justify the terms and conditions of their association (Cohen, 1989). In this type of society, citizens have a shared commitment to solve problems related to collective decision-making through public reasoning; they view the fundamental institutions of their society as legitimate becausetheyprovideaframeworkforfree and open public deliberation (Cohen, 1989). Simply put, “democracy is about the exchange of arguments and reasons for the common good, rather than contracts and negotiations between self-interested parties” (Landemore, 2023). This section will explain and argue that deliberative democracyandsoritionmethodsarethesuperiorwaytogovernAI.

At its core, AIRA’s design promotes the exchange of ideas. AIRA’s deliberative assembly encourages members tolearnfromexpertsandconsiderdiverseperspectivesfromtheirpeers.Its underlying logic is that the aggregation of ideas creates a collective intelligence of the group (Landemore, 2023). Given that global citizens see different parts of reality,togethertheyforma richer and more correct picture of the world. Considering the nuances and complexities of AI regulation, by fostering a dynamic and adaptive decision-making process, members of the assembly will bring different pieces of information and arguments, and guide each other to the bestsolution.

Many contemporary legislatures were initially designed with the intention of exchanging ideas and information such as in parliamentary debates. However, this is rarely the case in practice, due to the dominance of elites who are divorced fromlivedexperiencesonthegroundandskew debates. This is why AIRA combines deliberation and sorition methods, hand in hand. In addition to deliberation, sorition promotes inclusivity, while reducing influence of special interests.

AIRA contends that inclusivity is crucial because a diverse and inclusive assembly is better equipped to identify risks, unintended consequences, and ethical considerations than a group of homogeneous decision-makers (Hong, 2004). To achieve this goal, AIRA will adopt a sorition scheme. Sortition ensures a more accurate representation of a population, creating a decision making process that lends itself totheneedsandconcernsofpeoplefromvariouscultural,social and economic backgrounds. Compared to alternative methods of selection – elections or

11

appointments – the randomization elements in sortition have thehighestlikelihoodofproducing a sample that is representative of the true given population (Hong, 2004)(Clementetal.,2013). This is particularly important when considering AI regulation because AI technology has a far-reaching, and undiscriminating impact that transcends many national boundaries. Similarly, whereas traditional election processes or appointments by the political elite often result in decision-making powers concentrated by a few groups or individuals, sorition ensures the distributionofpoweracrossallgroups.

AIRA’s design of randomselection–asopposedtoelectionsorappointments–willdiminishthe influence of special interest groups, such as tech corporations or political factions. By eliminating career politicians, it reduces the likelihood of members forming long-term relationships with special interests. Since the terms would be shorter compared to traditional legislatures itwouldalsolimitthetimeavailableforexternalpressurestodevelop,whilethelack of election and campaign financing issue cuts off a key avenue for potential influence. These factors collectively foster a more independent decision-making process within the assembly Consequently, a randomly selected, deliberative assembly would be less susceptible to lobby, corporate or political pressure, and wouldbeabletodevelopanddeploytheregulationsitdeems suitable (Verba et al., 2018). It will also level the playing field among other stakeholders –smaller tech companies and under-represented ideological and demographic groups – thus creatingamoreinclusive,competitiveandinnovativeAIecosystemthatbenefitseveryone.

Despite the fact that governance by sortition allows room formoreinclusivity,diversity,andthe minimization of biases, some might argue that such a form of governanceposessomerisks.For example, with a random selection, participants are not checked if theyhavebackgroundsthatfit a certain role, so sortition can sometimes allow for a lack of expertise amongst the selected participants (Sustensis & Czarneck, 2019). However, the random individual represents exactly what sortition aims to accomplish, allowing for the most common of citizens, no matter where they are from or whattheyareeducatedin,toparticipateinagovernmentcommitteeresponsible for impacting and changing legislation that ultimately impacts their own lives. The fact that sortition includes the voice of those often marginalized and unheard in conventional political systems make sortition even more legitimate in truly representing the voices of the people. Furthermore, cognitive diversity enables the power of collective intelligence or collective wisdom,whichoftenoutperformsaperceivedlackofexpertise(Pecaric,2017).

Other detractors might still challenge that sortition leads to a lack of accountability from assembly members, causing a loss of credibility and distrust amongst the group (Sustensis & Czarneck, 2019). While members are not accountableinthespecificsenseofvotesandelection, there is nonetheless a more powerful accountability at play. The forum of public deliberation allows formoretransparencyandaccountabilityinthesenseoflegitimatepublicreasoning.With the lack of biases, conflicts of interest, power dynamics,alongwiththelargeamountofequality

12

that is promoted, sortition in turn is actually more accountable and transparent than most government styles. As a result, sortition is the most beneficial and efficient form for the AIRA despitetheconflictsitmaypose.

Some examples of sortition based legislatures are the citizens assemblies chosen by lot used in France and Ireland as a viable approach to governance. Both countries adopted a form of a citizens assembly in order to gain perspective on different issues facing the general population andcountryasawhole.

In 2023,theFrenchgovernmentdecidedtopursueacitizensassemblytoaddresstheissueofend of life or assisted dying. They used the method of sortition in order to choose the citizens assembly, which consisted of 170 participants compiled according to the criteria of age,gender, education, place of residence and occupation in such a waythatitrepresentedareflectionofthe population at the time with the youngest participant being 20 years old, and the oldest being87 years old (Burgerrat, 2023). The assembly met on weekends over an extended period of time, ultimately coming up with a multitude of ideas and proposals. Although not much legislation was implemented directly from the assembly and theprocessdidnotcomewithouthardshipand conflict, a group of citizens chosen at random were able to make a difference and showtherest of the country that this method of governance would be more useful in the future as it gave a voice to the overall population, which they felt, they otherwise had not had. Some members of the governance committee, in charge of overseeing the convention, had similar viewpoints. For example, Helen Landemore suggested that employing randomly selected deliberative bodies consisting of a representative cross-section of a country's total population to determine legislative programs and provide recommendations on important national issues results in democracies thataremoreinclusive,allowingforthecommoncitizentohaveavoice,promoting inclusivity and representation, thus allowing the government tofunctionmoreefficientlyoverall (Cummings,2022).

Additionally, in 2016, Ireland took the initiativetocreateacitizensassemblyinordertodiscuss, debate, and combat certain issues amongst the Irish society such as marriage equality, abortion rights, climate change, and others. The assembly consisted of 99 citizens who were chosen randomly, of whom met on weekends until the final meeting in April of 2018,andwereguided by experts available for consultation. The Irish citizen assembly managed to get two constitutional amendments passed dealing with marriage equality and their eighth amendment (Observatory of Public Sector Innovation, 2016) Similar to the French, the Irish were able to create a system of deliberative democracyinwhichthecommoncitizenofapopulationwasable to have a voice, rather than just the elected representatives and officials. Selection by lot inthis case enabled for more inclusivity amongst the citizens of Ireland, which in turn allowed for majorlegislativechanges,inwhichthegeneralpopulationfeltheard.

13

Sortition based legislature and citizens assemblies are becoming significantly more popular around the world. What was cited are but two examples in “a new wave of deliberative democracy,” where sortition based deliberative assemblies are being employed in city, state or national governance (Chwalisz, 2019). In each example provided, both countries helped to illustrate the inclusivity and representation sortition based governance allows for. More voices are heard, more individualsfeelrepresented,anditallowsforamoreefficientlyrungovernment, producing legislation on intractable or controversial topics. AIRAplanstousethephilosophyof sortition to promote inclusivity and efficientpractices.Theelementofrandomizationinsortition has the highest likelihood of producing a sample that is representative of the true global population. This isspecificallyimportantwhenconsideringAIregulationbecauseAItechnology has a global impact that exceeds national boundaries, which is why AIRA plans to use the philosophy of sortition based governance to promote inclusivity, representation and efficient, trustworthypractices.

These deliberative and sortition-based elements make AIRA a legitimate body for several reasons. First, not only is sortition the only way to foster democratic representation that is representative of the global population, but also its collaborative nature will promote a sense of global cooperation and shared responsibility (Cohen, 1989). Ultimately, these methods will establish trust among nations and stakeholders and increase the likelihood of compliance with the frameworks developed by the AIRA. Second, AIRA’s framework fosters a high level of transparency and accountability The reduced influence of political elites and powerful corporations due to random selection will ensure that AIRA’s decisions are best for the global population, not a small group of people – no more lobbying and corruption. Moreover, the selection of representatives through randomization creates a dynamic of equal opportunity for participation (Cohen, 1989). AIRA will be governedbyeverydaycitizens–thoseaffectedbyAI –ratherthancareerpoliticians.Ultimately,thisdesignminimizesbiasesandconflictsofinterest.

14

SECTION1.4:Jurisdiction

1.4.1:StandardsofRegulationandPurviewonCompaniesinCyberspace

Recognizing the increasingly common call for regulation of AI, the continued use of AI by the people, and the inevitability of AI regulation (Smuha, 2019), a demand for a diverse and collective modality of regulation for companies operating in cyberspace to be responsible, ethical, and accountable in their actions and operations is present.

Compliance with AI Infrastructure Security: The AIRA will recommend regulation for companies to take steps to protect their networks, systems, and data from cybersecurity attacks byhackers,cybercriminals,andothermaliciousactors.

Compliance with Consumer Regulation: The AIRA will recommend regulation for companies to protect and use inaccordancewithrelevantlawsandregulationstheirpersonaldatafromtheir customers and users, which can include sensitive information such as financial data, health information,andpersonalidentification.

Compliance with Ethical Considerations: The AIRA will recommendregulationforcompanies to navigate a range of ethical considerations related to the use of AI, automation, and other technologies and ensure that these technologies are used in ways that are fair, transparent, and accountable.

Compliance with Data Security Regulation: The AIRA will recommend regulation for companies tocomplywitharangeoflawsandregulationsrelatedtodataprotection,privacy,and cybersecurityandmeettheirobligationsundertheselawsandregulations.

Compliance with Social Norms: The AIRA will recommend regulation for companies to navigate and ensure that they are respecting social norms and contributing to a safeandhealthy online environment to build trust with and operate in ways that are culturally and socially acceptabletotheircustomersandusers.

Compliance with Accessibility for All: The AIRA will recommend regulationforcompanies to conduct themselves in a manner that encourages and maintains cyberspace for the good of all usersbyenhancingtheaccessibilityandusabilityofAItechnologyforeveryone.

Compliance with Humanitarian Values: The AIRA will recommend regulation for companies to conduct under principles ofinclusivegrowth,sustainabledevelopmentandwell-being,human centered values and fairness, transparency andexplainability,robustness,securityandsafetyand accountability(WEF,2021).

15

Compliance with Guidance: The AIRA will recommend guidance for consideration by policy makers with the purpose of maximizing and sharing the benefits from AI, whileminimizingthe risks and concerns, with special attention to international cooperation and inclusion of developingcountriesandunderrepresentedpopulations.

Compliance with Protection: The AIRA will recommend regulationforcompaniestorecognize the need to continue to promote the protection of privacy and personal data consistent with applicableframeworks.

Compliance with Capacity Development: The AIRA will recommend regulation for companies torecognizetheneedtopromoteAIcapacitybuildingandskillsdevelopment.

Compliance with Cooperation: The AIRA will recommend regulation for companies to strive for cyberspace cooperation and endeavor to work together with appropriate areas such as research and development, policy development and information sharing through open and collaborativeefforts(Franki,2023).

1.4.2:DevelopinganEthicalFrameworkfortheRegulationofAICreationandUsage:

Recognizing AI's potential for growth, empowerment, and shared prosperity, as well as its ability to transform society, and the need for ethical, responsible, and equitable AI development and application. This framework guides corporate policymakers, industry leaders, and citizens through the challenges of emerging technologies and their global impact while remaining mindful of potential risks and unforeseen consequences from neglecting ethical, social, and moral responsibilities.

Human-centered values and fairness: The AIRA will work toward a human-centric future society and emphasize the importance of working with all interested parties and stakeholders in sharing good practices and experiences (G20, 2019). TheAIRAwillcooperatewithgovernment regulation for actors developing AI-applications. The AIRA will respect the rule of law, human rights, and democratic values throughout the AI system lifecycle, including freedom, dignity, autonomy, privacyanddataprotection,non-discriminationandequality,diversity,fairness,social justice, and internationally recognized labor rights. The AIRA will implement mechanisms and safeguards that are appropriate to the context and consistent with the stateofart(OECD,2019). The AIRA will work closely with actors to prepare for and assist in the transformation of the world of work and of society by building human capacity. The AIRA will empower people to effectively use and interact with their AI systems across the breadth of applications. The AIRA will take steps toensureafairtransitionforworkersasAIisdeployed.TheAIRAwillalsowork closely with stakeholders to promote the responsible use of AI at work, toenhancethesafetyof workers and the quality of jobs, to foster entrepreneurship and productivity, and aim to ensure thatthebenefitsfromAIarebroadlyandfairlyshared.

16

Digital Transformation of Industries: The AIRA will share good practices and experiences regarding digital industrial policy with a view to promoting high quality inclusive development ofallsectorsandtakeactionstocreateafavorableenvironmentglobally.

Ensuring Functionality: The AIRA must be accountable for the proper functioning of its AI systems and for respecting the above principles based on its roles within a givencontextandbe consistentwiththestateoftheart.

Adaptation of AI Policies: The AIRA will review and adapt, as appropriate, their policies and regulatory frameworks and assessment mechanisms as they apply to AI systems to encourage innovationandcompetitionfortrustworthyAI.

Accountability: The AIRA must continue to address challenges related to privacy, data protection, intellectual property rights, and security to facilitate data free flow, strengthen consumer andbusinesstrust,andharnesstheopportunitiesofthedigitaleconomybycooperating toencouragetheinteroperabilityofdifferentframeworksandtheroleofdatafordevelopment.

Promoting Digitalization: The AIRA will support initiatives aimed at the promotion of investment in domestic and international digital connectivity infrastructure and connectivity technologies to avail connectivity to a greater number of individuals while promoting universal and affordable access to the Internet by all people by promoting connectivity in rural areas for ruralprosperity.

Fostering a Digital Ecosystem for AI by Investing in AI Research and Development: The AIRA will foster the development of a digital ecosystem fortrustworthyAIthatincludesdigital technologies, infrastructure, and mechanisms for sharing AI knowledge, and encourage public and private investment in research and development, including interdisciplinary efforts, to spur innovation in trustworthy AI that focuses on challenging technical issues and AI-related social, legal, and ethical implications and policy issues. Additionally, the AIRA will consider public investment and encourage private investment in opendatasetsthatarerepresentativeandrespect privacy and data protection to support an environment for AI research and development that is freeofinappropriatebiasforimprovedinteroperabilityanduseofstandards(FLI,2017).

Digital Literacy: The AIRA will provide users with the right information on the skills and knowledgetointeractwithitsAIsystems.

Bridging the Digital Gender Divide: The AIRA will reaffirm that the participation of historically marginalized gendercommunitiesinthedigitaleconomysupportsstrongereconomic growth, innovation, and inclusiveness and enhances societal well-being, and will encourage countries to take actions to bridge the digital gender divide, including through the development of frameworks to measure and track sex-disaggregated data, and to make efforts to increase historically marginalized gender communities’ accesstodigitalnetworkswhileaddressingabuse

17

and violentonlinebehavior,enhancehistoricallymarginalizedgendercommunities’participation in AI, and support historically marginalized gender communities’s entrepreneurship in digital businessandworkwithinexistingpartnershipsandframeworks.

Inclusive Design for People of all Ages/Persons with Disabilities: The AIRA will promote digital literacy strategies with a special focus on vulnerable groups for labor market transformation through developing and adapting digital technology to be user-friendly and human-centered for use by a diverse group of persons, including persons with disabilities,older persons,orthosewithlowerdigitalskills.

Building Human Capacity, Preparing for Labor Market Transition, Inclusive Growth, Sustainable Development, and Well-Being: The AIRA will promote international trade and investment as important engines of growth, productivity, innovation, job creation, and development for creating an inclusive,sustainable,safe,trustworthy,andinnovativesocietywith digitalization, and will work towards sustainable development goals through the increasing convergenceofthephysicalworldandthevirtualworld.

Creating a Beneficial Consumer Environment: The AIRA will create a consumer oriented digital environment that has a direct impact on the ability of people to reap the benefits of digitalizationfortheirpersonalandprofessionallife.

Regulation of Innovative Technologies: TheAIRAwillstriveforinnovation-friendlypoliciesto capitalize the potential of digital technologies and look to remove barriers to innovation accordingly.

Standards for Regulatory Cooperation: The AIRA will conduct itself with interoperable standards that align with international and national stakeholders to achieve regulatory cooperation.

International Co-operation for Trustworthy AI Development: The AIRA will actively cooperate in the global and regional cyberspacespherestoadvancetheseprinciplesandtofoster the sharing of AI knowledge and progress on responsible stewardship of trustworthy AI. The AIRA will work together with theglobalandregionalforatofosterthesharingofAIknowledge and encourage international, cross-sectoral and open multi-stakeholder initiatives to garner long-term expertise on AI and promote the development of multi-stakeholder, consensus-driven global technical standards for interoperable and trustworthy AI while encouraging the development, and its own use, of internationally comparable metrics to measure AI research, development and deployment, and gathering of evidence to assess progress in the implementationoftheseprinciples.

Developing of Smart Cities: The AIRA will encourage networking and experience-sharing for the growth and development of sustainable and inclusive smart cities and urban growth areas

18

where most of the world’s population and energy consumption are concentrated that take into account transparency, resiliency, privacy, security, efficiency, and interoperability by sharing good practices and lessons learnt from its experiences in solving social problems through differentdigitalregulatoryframeworks(U20,2018).

International Co-operation for Confronting Societal Challenges: The AIRA will work along international and national policymakers to address a wide range of societal challenges and facilitate discussion on how technology can be better incorporated into policy tools to not only beinnovation-friendlybutalsoinnovativeitselfwithoutlosinglegalcertainty

Transitioning from R&D to Deployment and Operation: The AIRA will promote a policy environment that supports an agile transition from the research and development stage to the deployment and operation stage for trustworthyAIsystemsandconsidersusingexperimentation toprovideacontrolledenvironmentinwhichAIsystemscanbetestedandscaled-up.

Promoting AI Security in the Digital Economy: The AIRA will provide and facilitate an ongoing discussion of security in the rapid expansion of emerging technologies, including the Internet of Things and the growth of the digital economy, to raise awareness on the importance ofactionstoenhancesecurity.

Resisting Malicious Actors and Behaviors: The AIRA will reaffirm its commitmenttofighting the exploitation of the Internet for violent extremist and terrorist purposes by promoting a free, open, and secure internet and encouraging the digital industry to continue working with all stakeholders to fight the use of the internet and social media for violent extremist and terrorist purposesandaddresscontentthatincitesterroristacts.

Localized Security: The AIRA will develop localized and customized frameworks and methodologies that recognize the global aspect of security in the digital economy by following industry-led and market-led global technical standards, developed based upon principles of openness, transparency, and consensus to help deliver interoperability and promote trust by recognizing the role played by stakeholders such as the private sector, the technical community andcivilsociety,andrelevantinternationalorganizations.

Robustness, Security and Safety: The AIRA will promote traceability, including in relation to datasets, processes and decisions made during the AI system lifecycle, to enable analysis of an AIsystem’soutcomesandresponsestoinquiry,appropriatetothecontextandconsistentwiththe state of art and based on its roles, the context, and their ability to act, apply a systematic risk management approach to each phase of an AI system lifecycle on a continuous basis toaddress risksrelatedtoAIsystems,includingprivacy,digitalsecurity,safetyandbias.

Transparency and explainability: The AIRA will commit to transparency and responsible disclosure regarding AI systems to foster a general understanding of AI systems by making

19

stakeholders aware of their interactions with AI systems, including in the workplacethatenable those affected by an AI system to understand the outcome (OECD, 2019) ans to allow an actor adversely affected by a corporate AI entity to challenge its decision or outcome based on plain and easy-to-understand information on the factors, and the logic that served as the basis for the prediction,recommendationordecision.

Data Free Flow with Trust: The AIRA will commit to cross-actor flow of data, information, ideas, and knowledge to generate higher productivity, greater innovation, and improved sustainable development by recognizing the sharing of good practicesandutilizingtheavailable list of digital policies other than and including open source data systems and will consider promotingmechanismstosupportthesafe,fair,legalandethicalsharingofdata.

1.4.3:CooperationwithNewandExistingOrganizationsandRegulatoryStructures

Recognizing the common interest by countries to not only build but also export AI products and services, and corporate interest in capitalizing on these products and services, as well the common aim to ensure the minimisation of the technology’s harm to citizens.

The AIRA will endorse and join organizations aiming to enhance transparency on regulatory regimesrelatingtoAI.

The AIRA will cooperate with existing industry and state organizations to strategically take the leadinAIstandard-setting.

1.4.4:EnablingandEncouragingTransparencyandParticipation

Recognizing the disparity of the pace of innovation to the pace of regulation and that regulatory systems of coopetition for artificial intelligence which acknowledge that concrete and binding rules tied to an effective enforcement mechanism for actor deviance will provide more stability than vague or voluntary guidelines, and that even the most binding standards and agreements can be breached, especially in areas where actor sovereignty is deemed a priority

The AIRA will enable transparency in their agenda setting through protective regulation or guidelines that set out how AI can be used ethically in a specific sector or use case (Vinuesa, 2020).

TheAIRAwillrecommendregulationforcompaniesdevelopingAI-applications.

The AIRA will provide to the local regulatory arena the economic incentive of the ‘first mover advantage’(Belton,2019).

20

The AIRA will retain thepossibilityandincentiveforindividualactorstoadoptahigherlevelof protectionthanthecommonlyagreedstandard.

The AIRA will condone the possibility to impose sanctions coupled with pressure from the national, international and cyberspace community against companies to comply to decrease the chanceofstandardandagreementviolation.

The AIRA will favor an approach in which it can learn as it goes and have the spacetotestout different and competing models of regulation to ultimately lead to a discovery of the best regulatory results in a manner that provides affordable testing time where the costs attached to non-interventionortogettingitwrongarenotprohibitive.

1.4.5:EstablishingModalitiesofRegulation

Recognizing the need for a diverse set of regulatory approaches to reflect the diverse economic, social, legal and political situations of the world, the differing manner in which countries will be affected by AI and the necessity of an even distribution of wealth among states affected by the global boost in economic growth enabled by AI.

The AIRA will recommend regulation for developing AI-applications, regulation for AI developers and deployers, or guidelines for ethical AI use in a specific sector or use case (Custers,2023).

1.4.6:AgendaSetting

Recognizing the need for an agenda-setting system for effective functioning and goal achievement that promotes focus, productivity, efficient time management, communication, collaboration, inclusivity and encourages goal-setting, following-up, transparency, and accountability while ensuring that organizations remain aligned with their objectives and fostering a culture of trust among stakeholders.

This AIRA will designanagenda-settingprocessthatiscomprehensive,structured,andeffective inaddressingitsdiversityofconcerns.

TheAIRAwillidentifythekeyconcernsfacingtheAIRAthroughthreatintelligence.

TheAIRAwillprioritizeconcernsbasedontheirpotentialimpactontheAIRA.

The AIRA will set clear goals and objectives that align with the AIRA’s overall strategy and objectives.

21

The AIRA will establish a schedule for its agenda that maximizes the availability of key stakeholders(Smith,2000).

The AIRA will assignspecificresponsibilitiesforeachagendaitemthatensurethatallissuesare addressedandthattheAIRA’sgoalsareachieved.

The AIRA will engage stakeholders, including staff, senior management, community and board members to ensure that their perspectives and input areconsideredintheagenda-settingprocess (Smith,2000).

The AIRA will regularly review and update their agenda to ensure that it remains relevant and alignedwiththeAIRA’sneedsandobjectives.

22

PART2

SECTION2.1:AIRAStructure

The Artificial Intelligence Regulation Authority (AIRA) will be composed of two chambers—a GeneralAssemblyandaGovernanceCommittee.

TheGeneralAssemblywillbea1170personin-personad-hoclegislativeassembly,composedof randomly selected citizens from across the globe. This assembly will be a rotating body, with eachcohortactivefora2-weekintensive.

Since the travel and time of the participants will be compensated for, we expect assembly members to participate in programming and deliberation for 7 hours a day with breaks. For reference, the Global Assemblyspent“68hourstogetherover11weeks”intheirprocess(Global Assembly, 2023). By increasing the total deliberative time and inviting participants to join in-person,wehopetohaveabetterexperienceforparticipantsacrosstheboard.

Furthermore, by ad-hoc, we mean that the assembly might convene at irregular intervals, depending on the regulatory demands of the public and the agenda-settingprocess.Thepurpose of an ad-hoc body is to give participants an opportunitytofullydedicatethemselvestothetopic of Artificial Intelligence without interruptions that would be associated with a longer time commitment. The main responsibility of the General Assembly is to propose and develop legislationonthegovernanceofAI.

The second chamber, the Governance Committee, will be a smaller 25 person body tasked primarily with the implementation of the principles herein stated and regulations passed by the General Assembly. This bodywillhavethepowertodeclarecompaniesincompliance,andissue sanctionsandpunishmentsforthosewhoviolate.

Lastly, AI disputes arising from these regulations will be resolved through a system of international arbitration, which will legally bind and hold accountable all AI companies who submit to AIRA'sjurisdiction.Whilenotapermanentbody,thismechanismwillbecrucialtothe functioningofAIRA.

Section2.1.1GeneralAssembly(Legislative)

What is the purpose of this Assembly?

The purpose of this assembly is to gather citizens from across the world and assess what the impact of AI willbeontheireverydaylives,communities,andbeyond.AsArtificialIntelligence touches nearly every aspect of our lives, AIRA takes the position that everyone is a relevant

23

stakeholder. Whether one feels the impacts of AI through one’s job, residence, or community, everyone’s voice should be included in the conversation surrounding its future. Specifically,the roleofAssemblymembersareasfollows:

(1) Propose or amend constitutional framework to best address ethicsorconcernsontheuse ofAI

(2) RespectcrosscountryrelationsandattitudestowardsAI(softpower)

(3) ConsultwithexpertsonAItodeterminereasonableandresponsibleregulations

What is the composition of the Assembly?

This regulatoryassemblywillbecomposedofcitizensfromaroundtheglobe.Byopeningupour selection process to everyone, we ensure that no one can be excluded on the basis of their job, economic standing, education, or otherwise that would be meaningfully affected by issues of artificialintelligence.

This General Assembly willbecomposedof6representativesfrom195countries.Forreference, theUNcurrentlyhas193membercountrieswiththeexceptionofTaiwanandKosovo.1

Our selection mechanismwillmimicthatusedintheGlobalAssembly’sCoreBodyonClimate.2 Specifically, using the NASA database of human population density and a process of sortition, six geographic points will be randomly drawn across eachcountrywhereourteamwillvisitand recruit 1 individual in that city or neighborhood. Sortition is followed at each level to limit the bias our field teams may introduce and to ensure that we have a representative body at the internationallevel.Additionally,participantswillconsenttothisprocessentirelythroughout.

How will this Assembly function?

Members of this assembly willbeaskedtoconveneforanintensive2weekstolearnanddiscuss issues relating to AI. They will spend time meeting with experts, relevant stakeholders, and reviewing existing legislation. Following this, the assembly will be asked to review key principles and concerns into a single framework to regulate AI. The constitution herein is amendable and will be the starting point for this group. The final product will then be voted on andratifiedby2/3ofallparticipatingmembers.

1 Currently,theUNdoesnotrecognizeTaiwanorKosovoassovereign;rather,aspartofChina andSerbia,respectively.WithrespecttoTaiwanspecifically,ifourbodyhopestohaveanyform oflegitimacyandinfluenceoverChineseprivatecompanies,asignificantlylargepartoftheAI “market,”wemaybeforcedtotakeasimilarpoliticalstance.

2 https://globalassemblyorg/about-2

24

Assembly members will be compensated for their time withasalaryconsistentwiththoseofthe UN Compensation and Classification Section.3 Members will serve single terms as part of this assembly, rotating as to give an opportunity for others to participate and not serve as a political careerbuildingopportunitypolitics

What are the obligations of the Assembly?

A number of things will be required of Assembly participants. They will be responsible for studying this issueandembracingtheirdiscussionopportunities.Assemblymemberswillalsobe askedtoengagethoroughlyindiscussionsandgroupdebatesledbyAIRAfacilitators.

What is the role of Facilitators

Recognizing that mass deliberation may favor those with the loudest voice, our facilitators are there to ensure everyone’s voice is heard equally and consistently throughout the deliberation process. This is the main responsibility of our facilitators to the General Assembly While the role of facilitators requires them to inhabit the same space as assembly members on the legislative floor, facilitators are not to have a role in voting or proposing legislation. Toaddress potential biasorimproperbehavioronthepartofgroupfacilitators,assemblymemberswillhave powertovoteandreplacetheseindividualswithasimplemajority

Facilitators will be randomly selectedfromapoolofscreenedprofessionalswhohaveworkedin similar deliberative roles. Recognizing that facilitators in other deliberative democracy projects, such as in Ireland and France, may not have been entirely representative of their country’s population,ourgoalistobuildonthoseexperiencesandstriveforbetter.

What is the role of Experts

The role of experts in our assembly is to strictly offerinsightintotheirrelevantAI-relatedfield, and distill technical information into accessible information. Experts are invited to fully participate in discussions and give their opinions and thoughts on legislation, but are unable to vote. The reason for this is two-fold. First, as described by political philosopher, Hélène Landemore, in her book Open Democracy, the benefit of giving experts an advisory role rather than a decisional one is to maintain the integrityofopeningupdemocraticprocessestoordinary citizens. Put simply, the goal is to have “experts on tap,notontop,”asLandemorerecommends (Landemore, 2020). Secondly, restricting the voting powerofexpertsalsoallowsAIRAtoavoid instances of private-sector representatives voting on potential legislation that may be profitable forthem.

In the end, we maintain that inviting experts to join the Assembly in conversation is still important to our deliberative process. There is no reason to believe that exposure to the Assembly’s conversation won’t have a meaningful impact on the broader AI field, especially if

3 https://wwwun org/Depts/OHRM/salaries allowances/salaryhtm 25

expertsareexposedtotheviewsofanassembledpublic.Elevatingthevoicesofcitizensisacore aspectofAIRAandthissituationisnodifferent.

Experts will be chosen based on nominations from companies, research institutions, and universities. As we will require these individuals to offer specific expertise on topicswithinAI, it'simportantthattheseindividualsbeleadersintheirrespectiveresearchorbusinessfields.

What is the role of the Assembly to companies?

The General Assembly ought to be inclusive totheperspectiveofthecompaniesitisworkingto regulate. Recognizing that the challenge ahead requires an immense amount of global cooperation,itisessentialthatourAssemblybeledbyexample.

At the beginning of the Assembly’s convention, companies will be invited to voice their thoughts, opinions, and concerns on what they believe should be the direction of AI regulation. Followingthisopenforum,theGeneralAssemblywillconsiderinputfromourcompanypartners in the form of written communication. WhileitisnotguaranteedthattheAssemblywilltakethe desired directionofcompanies,itisessentialthatthosemodesofcommunicationbeopeninboth directions.

What is the role of the general public?

The general public is undoubtedly an important aspect of any legislative-making process. The public is the ultimate stakeholder in any legislation passed by a government or regulatory body such as AIRA. Legislation often has far-reaching effectsonsociety,andthegeneralpublicisthe groupthatismostlikelytobedirectlyimpactedbyit.

Accordingly, people outside of our Assembly will have the opportunity to participate in public forums–digital and in-person–to make known their concerns and thoughts abouttheAssembly’s proceedings. AIRA will strive for transparency at every logistically-feasible level,andourwork will reflect that goal. In addition to public forums, the general public is encouraged to engage withtheirexistinglocalorStategovernmentsandadvocateforstrongerAIregulations.

Section2.1.2GovernanceCommittee(Executive)

What is the purpose of the Governance Committee?

The main purpose of this body is to defend the principles and regulations of AIRA. The Governance Committee (GC) has the power to declare companies in compliance, and issue sanctionsorpunishmentsforthosewhoviolate.

This body is unable to propose legislation. They should also be required to hold distance from theGeneralAssemblyunlesscalledupontotestify

26

What is the composition of the Governance Committee?

The second chamber of AIRA, the Governance Committee, will be a smaller 25 person regulatory body. As the body responsible for overseeing AIRA regulations, this body’s smaller size reflects one capable of making decisions at a faster rate than the General Assembly. Empirically, the size of cabinets around the world is cappedat40members,withmostcountries in the range of 20 to 30 members (Taagepera et al., 2018), so AIRA follows a similar example. Members of the GC, due to their required expertise, will not be chosen by lot, but rather by nominationtoandconfirmationbytheGeneralAssemblybya2/3thsvote.

How will legislation passed by the assembly be enforced?

Depending on the legislation, companies will be required to provide evidence for their compliance with given legislation. This may be in the form of structuredfilingsandreportsthat may be audited by the general public. Compliance may be rewarded as stipulated in the legislation. Violations or non-compliance, conversely, will be addressed with punishments such as fines as stipulated in the legislation. Repeated infractions will result in a dispute between AIRA and the infringing company that will be settled through international arbitration (see Section2.3Judiciary).

What is the Governance Committee’s responsibility to companies?

GC owes as muchtransparencytocitizensasitdoestocompanies.Especiallywithanagreement as delicate as this, it's important that members see themselves and their values represented by thisbody.

SECTION2.1.3:InternationalArbitration(Judiciary)

What is the judiciary mechanism?

Disputes around non-compliance with respect to AIRA legislation will be resolved through internationalarbitration.Internationalarbitrationisaformoflitigationthattakesplaceinfrontof private arbitrators instead of in domestic courts (Aceris Law, 2021). It is already the preferred litigation method for many businesses around the world. Arbitration awards are recognized and enforceable internationally in 172 countries in the world thanks to the New York Convention (United Nations Convention on the Recognition and Enforcement of Foreign Arbitral Awards, 1958)(NewYorkArbitrationConvention,2023).

27

The exclusion of some countries shall not prevent the enforceability of arbitration awards. Firstly, it is still possible to pursue enforcement on the violating company through another country that is party to the New York Convention. For example, Libya isnotapartytotheNew York Convention, so it is not currently possible to enforce an arbitration award in the state of Libya. However, if the AI company also operates in any oneofthe172countriesthatispartyto the New York Convention, then the same arbitration award can be enforced in any of those countries. Secondly, the case where AIcompaniesonlyoperateincountriesnotpartytotheNew York Convention in order to avoid enforceable legislation face the disadvantage that consumers know with high plausibility they have suspicious compliance to AIRA legislation. Contracting states for the New York Convention are also still growing such as Iraq joining in 2021 and we anticipatemorecountriesbeingabletoenforceAIRAlegislationasaresult.

What is the composition of the judiciary body?

The judiciary body will be a tribunal of arbitrators. As is common practice in international arbitration, disputing parties nominate an arbitrator each, and the two nominated arbitrators jointly nominate a head arbitrator; the three arbitrators then become the tribunal for the dispute (ICC Rules of Arbitration, 2021). Note that this means the tribunal of arbitrators differs dependingondispute,unlikeapermanentcourtorsetofjudgesinadomesticsetting.

How are issues raised to the judiciary body?

Disputes will be raised as in any lawsuit, and the body of arbitrators will form according to the internationalarbitrationrules.

Why international arbitration?

Firstly, international arbitration allows AIRA to not need for domestic legislators to integrate AIRA legislation in order for them to be enforceable. Given the varying degrees and different legislative processes in different states, international arbitration allows AIRA legislation to be applied equally and concurrently across all countries thanks to international arbitration’s cross-borderquality.

28

Secondly, international arbitration removes the need for a fixed court. Not only does this approach save cost, it also sidestepstheselectionbiasorsuspicionsofnon-neutralitythatplague permanentinternationalcourtsliketheInternationalCriminalCourt(Kimenyi,2013).

Finally, international arbitration is already the preferred mode to settling disputes compared to traditional court litigations in many settings for companies, for its speed and cost-efficiency (Aceris Law, 2021). Since AIRA regulates companies, it will be a familiarandamenablemeans ofsettlingdisputesarisingfromAIRAlegislation.

SECTION2.2:PotentialChallenges

Why one assembly instead of multiple national assemblies?

Having multiple national assemblies, though a nice sentiment, would be impractical. For one, it would require all 195 countries, of which only 24 countries are considered full democracies, to agree tohostandallowtheircitizenstoparticipateinademocraticcitizens'assembly Thisseems extremely unlikely and the inconsistencies in practice will likely impinge on the ability of assemblytomake robustglobalregulationswherenocountryfallsthroughthecracks.

There would also be the issue of synthesizing all 195assemblies’recommendationsintoasetof global recommendations. This would then require an additional global assembly hypothetically requiring a few representatives from each of the individual national assemblies. No matter the method by which such representativeswouldbehypotheticallychosen,itisunclearthattheycan adequately represent the decisions from the local level to the global level without overemphasizing their biases —this drop in representation hinders the legitimacy of such a two-tiered system. Thus, even though having multiple national assemblies might, in theory, produce a more globally representative product, it is an unrealistic goal that would never be feasiblegiventherestrictionsofthenatureofthedemocraticprocessinquestion.

Why a rotating body instead of a static body?

Members in this assembly will serve for two weeks before being rotated out for other representatives fromthesamecountry.Thisservestwoprimarypurposes.Firstly,thisalsoallows for the whole process to be more democratic in nature as it enables more people to participate, thus resulting in a higher probability of diversity. Additionally, full rotation was chosen over staggered rotation because it assures that every member will be there for an equal amount of time.Secondly,theprocessofrotationmakesparticipatingintheassemblymoreappealing.Ifwe were to randomly select six representatives from each country and request that they move to a new location —separated from their family and life for an extended period of time— there’s a high likelihood that many will dropoutmidway,hinderingtherepresentativenessandlegitimacy oftheassembly.Thisdoeshoweverraisetheissueofstability.

29

Since every two weeks the entire assembly will essentially be changed for a new group of people, one might argue that it is likestartinganeweverytime.Thishoweverisremediedbythe duration of each session and the continuity of the experts. The two-week time period allows roughly a week for members to get acquainted with their surroundings, familiar with how the assembly will operate, and study the content and nature of the issue being discussed. The following week will then primarily focus on deliberation with the experts at hand should the citizenswantclarificationonanobjectivequestion.

Why six citizens from each country?

Six citizens from each country create an assembly of 1170 individuals, allowing it to be representative of the entire global population. (For data scientists a minimum of 1,067 individuals is considered sufficient to represent a population as vast as the entireworldwithina 3%marginoferror(Bullen,2022)).

Why not choose descriptive representation (e.g. demographically reflecting current users, or proportional to the population size of each country)?

We choose to do equal representation to buttress the voices of the minority If we had done descriptive representation the conversation would largely be dominated by countries such as China (1.4 billion people), India (1.3 billionpeople),andtheUnitedStates(331millionpeople). Thus the resulting outcome of the convention would largely be skewed infavorofeachofthese respectivecountries'values.

Why not a global random sample?

Unlike the recent (2021) Global Citizens’ Assembly ontheClimateandEcologicalCrisiswhich chooses representatives using a global civic lottery system based on NASA population density, we decided to scale down to each sortition at the national level. By doing so, it ensures that all 195 countries have an equal say in the deliberation process. Each country also comes with its own unique culture and perspective on technology and it is important to make sure all such voices are heard and takenintoconsiderationwhendraftinglegislationthatwillbeenforcedona globalscale.

How do we know that these citizens can adequately represent the perspectives of all users?

This is admittedly one of the limitations of the construction ofthisassembly.Giventhefactthat there are only six members from each countrypresentduringasession,itisimpossiblethatthey will demographically represent the perspectives of all users in said country. However, we once again attempt to remedy this through the idea of rotation. Additionally, the fact there are six representatives from eachcountry(chosenbyrandomselection)increasestheprobabilitythatwe willobtainalevelofcognitivediversitythatwillresultinarepresentativeproduct.

Why is the Governance Committee not based on sortition?

30

The goal of the Governance Committee is to execute on the legislation passed by the General Assembly, the scope of their work being limited to auditing companies, pronouncing them in compliance or out of compliance, and meting out rewards and punishments as stipulated by regulation from the General Assembly. Thus, deliberation is notakeyaspectoftheirrole,rather executive efficiency is prioritized. The requirement for large numbers and random selection do not fitintothisexecutivepriorityunlikethedeliberativerequirementforthelegislativeassembly. The Governance Committee is nonetheless subjected to nomination and confirmation by the GeneralAssemblyandhencehasthevoteofconfidenceoftheGeneralAssembly.

31

SECTION3.1:Implementation

Why are companies and shareholders likely to agree to submit to AIRA?

Digital society must be built on trust among all stakeholders including governments, civil society, international organizations, academics and businesses through sharing common values and principles–including equality, justice, transparency and accountability–taking into account the global economyandinteroperabilityineacheconomy.The(non-)adoptionoflocalregulation of AI could leadtonegativeexternalities.Suchexternalitiesarisewhereregulatoryactions–ora lack thereof – by one actor makes another actor worseoff,whilethefirstactordoesnot(oronly marginally) bears the costs. In principle, if an actor decides against the adoption of protective regulation of AI, thenegativeconsequencesthereofwill(also)befeltwithinitsownjurisdiction. Accordingly, it would need to adopt regulationthatavoidsnegativeinternalitiesbysafeguarding therightsandwellbeingofitsowncitizens.Inessence,onecompanyadoptingAIRA’sregulation will encourage other companies to adopt AIRA’s regulations as well. If the other actors did not adopt AIRA’s regulation, comparatively those actors would appear less trustworthy and less compliantwithindustrystandardsthanthosewhoobservewithAIRA.

How does AIRA affect competing jurisdictions with different technological regulations?

The costs of moving AI resources from one jurisdiction to another (be it AI talent, capital or infrastructure) are unlikely to be negligible. Companies developing embedded AI products typically rely on large factories, but it can also hold true for non-embedded AI producers or service providers to divide labor among different jurisdictions in view of the massive energy infrastructures they require for data storage and analysis. Furthermore, legal obstacles might further complicate the possibility of moving jurisdictions. The ease of the re-domiciliation of companies, for instance, depends on the originating jurisdiction and destination jurisdiction. Some regulators aiming to attract AI talent consider adapting their migration policies to more easily allow experienced individuals to move to their jurisdictionofchoice,butpublicdiscourse in many countries is moving towards a call fortightening(im)migrationregimes.Thisrendersit more difficult for regulatory subjects to relocate talent without global accord on AI jurisdiction differences and the ensuing employee abilities. AI companies would benefit from having an internationally accepted code of conduct that allows for cooperation between jurisdictions, ultimately ensuring user privacy protections while simultaneously allowing for the ethical developmentofmoreadvancedartificialintelligence.

How will AIRA impact consumer psychology and willingness to use AI?

Consumers are more loyal to and willing to spend more on companies ensuring ethical AI. Customer trust –whetheritconcernsabusiness-to-businessorbusiness-to-consumercontext–is a prerequisite for the adoption of the technology Consumer trust can also be priced. Ethics and

PART3
32

competitiveness go hand in hand. Businesses cannot be run sustainably without trust, and there can be notrustwithoutethics.Andwhenthereisnotrust,thereisnobuy-inofthetechnology,or enjoyment of the benefits that it can bring. An ethical approach to AI strengthens citizens’trust indigitaldevelopmentandaimsatbuildingacompetitiveadvantageinAIcompanies.

As the General Assembly sources its legislation from customers and consumers themselves, AIRA will facilitate customer trust in a very beneficial manner for AI companies, augmenting company transparency and user participation. Due to this logic, several AI companies have previously stated their desire for a separate regulatory framework. For example, Meta representatives have said that the company wants to see “differenttypesofcouncilsorbodiesto talk about standards” to regulate AI corporations broadly (Levy, 2022). Having an external regulatory body supplies companies with ready-made answers, based on user experience and broad popular opinion across all sections of the world, for difficult questions on AI ethics and use. External directives remove the pressure from AI companies to come up with solutions themselves under broad scrutiny from shareholders, customers, and the world, especially when they are ill-positioned to do so, being divorced from the myriad lived experiences of users on their products around the world. AI companies will not have to deal with criticism for making hard choices on ethics if a third party oversight commission has the responsibility of creating rules instead of the companies themselves. Thepresenceofanexternalregulatorybodyremoves AI company culpabilityandeffectivelylendsvaliditytothesecompanies.Theglobalperspective that lottocratic deliberation ensures compounds AI companies’ legitimacy and will increase futureAIconsumption.

How will the Artificial Intelligence Regulation Authority respect existing AI regulation?

Built upon AIRA’s values of respecting the cyber sovereignty of individual countries and countries’ ability to determine their own laws of content moderation, AIRA will work in conjunction with countries and international stakeholders to ensure the development, implementation, and continued oversight of ethical AI policy for the protection of users. The existing framework guidelines, such as those prepared by the U.S. Department of Commerce, OECD, and the Beijing Artificial Intelligence Principles, will be used as materials to educate members of the General Assembly. Organizations like Partnership on AI that provide AI development guidelines may be askedtoparticipateinAIRAasanExpert,orhavetheirfindings bepartofthematerialsusedtobriefmembersoftheGeneralAssembly(PartnershiponAI,n.d.).

AIRA will not be antagonistic to existing regulatory frameworks. Rather, AIRA strives tobring into conversation these frameworks, experts, AI’s affected user base, and the world, to create viable solutions together. Since there is close to no binding regulation concerning AI implementation and development intheworld,AIRAwillfilltheinternationalgapinlegislation. By usinganinternationalperspectivetocreatethelegislation,includingprimarilytheopinionsof users, supplemented with experts, shareholders, anddevelopersthemselves,AIRAwillallowfor

33

all current guidelines and opinions on AI regulation to be synthesized into one cohesive codificationthatencouragesethicaldevelopmentandprotectsprivacy.

AIRA acknowledges that the inherent complexity of a globalized world with national diversity requires a deliberative model of regulatory cooperation that acknowledges the need for a combination of competition and cooperation. Thecommittee’sfocusshiftsawayfromadualistic choicetowardsinsteadidentifyingareasthataremoresuitedforaconvergingapproachandareas that would rather benefit from a diverse set of competing regulations. AIRA is committed to ensuring that the power potential created with AI is distributed equally. Because allareaffected by the internet, all are affected by AI. AI regulation should hinge upon collective intelligence, and all persons of the world deserve the opportunity to participate in the decision-making processes addressing AI regulation, development, and use. In the future, as it develops and is better understood, AI could even possibly be used to help pinpoint those areas, andhenceassist regulatorsintheirquestforanadequateregulatoryapproachtoAI.

34

Bibliography

Acemoğlu,Daron.(2021,November23). Dangers of unregulated artificial intelligence.CEPR. RetrievedApril24,2023,fromhttps://cepr.org/voxeu/columns/dangers-unregulated -artificial-intelligence

AcerisLawLLC.(2021,November26). What is international arbitration? .International Arbitration.RetrievedApril23,2023,fromhttps://www.international-arbitration-attorney. com/what-is-international-arbitration/

AlgorithmicAccountabilityActof2022,H.R.6580,117thCong.(2022).

https://www.congress.gov/bill/117th-congress/house-bill/6580/text

Anyoha,R.(2017,August28).TheHistoryofArtificialIntelligence.ScienceintheNews.

RetrievedApril17,2023,fromhttps://sitn.hms.harvard.edu/flash/2017/history-artifi cial-intelligence/

Beijing Artificial Intelligence principles InternationalResearchCenterforAIEthicsand Governance.(2022,January10).RetrievedApril11,2023,from

https://ai-ethics-and-governance.institute/beijing-artificial-intelligence-principles/

Belton,K.B.,Audretsch,D.B.,Graham,J.D.,&Rupp,J.A.(2019).WhoWillSettheRulesfor SmartFactories?Leadershipininformationgovernancewillprovideafirst-mover advantagetothenation'smanufacturingsector Issues in Science and Technology, 35(3), 70+.https://link-gale-com.yale.idm.oclc.org/apps/doc/A583489417/OVIC?u=290 02&sid=summon&xid=400a2cac

Bardot,F (2022).ArtificialIntelligence,RobotsandUnemployment:EvidencefromOECD Countries. Journal of Innovation Economics & Management,37,117-138.

https://doi.org/10.3917/jie.037.0117

Browne,R.(2023,April17). Italy became the first Western country to ban CHATGPT. Here's what other countries are doing.CNBC.RetrievedApril11,2023,from

https://www.cnbc.com/2023/04/04/italy-has-banned-chatgpt-heres-what-other-countries-ar e-doing.html

Browne,R.(2023,April17). With chatgpt hype swirling, UK government urges regulators to come up with rules for A.I.CNBC.RetrievedApril11,2023,from

https://www.cnbc.com/2023/03/29/with-chatgpt-hype-swirling-uk-government-urges-regul ators-to-come-up-with-rules-for-ai.html

Bullen,P.B.(2022,September28). How to choose a sample size (for the statistically challenged).PracticalToolsforInternationalDevelopment.RetrievedApril24,2023, fromhttps://tools4dev.org/resources/how-to-choose-a-sample-size/#:~:text=A%20 good%20maximum%20sample%20size,%2C%2010%25%20would%20be%2020%2C000 Burgerrat.(2023,April2).Frenchcitizens'assemblysupportsassisteddying.

https://www.buergerrat.d/een/news/french-citizens-assembly-supports-assisted-dying/ Burrell,J.(2016). How the machine ‘thinks’: Understanding opacity in machine learning algorithms.BigData&Society,3(1),205395171562251.

https://doi.org/10.1177/2053951715622512

Chwalisz,C.(2019).Anewwaveofdeliberativedemocracy. Carnegie Europe, 26,1-6.

35

ClémentRJ,KrauseS,vonEngelhardtN,FariaJJ,KrauseJ,KurversRH.Collectivecognition inhumans:groupsoutperformtheirbestmembersinasentencereconstructiontask.PLoS One.2013Oct17;8(10):e77943

Cohen,Joshua. Deliberation and Democratic Legitimacy,inTheGoodPolity:Normative AnalysisoftheState,AlanHamlinandPhillipPetit,editors,NewYork:Blackwell,1989.

Cummings,M.(2022,November29). Yale's Landemore helps guide France's unique citizens' convention YaleNews.RetrievedApril26,2023,from https://news.yale.edu/2022/11/29/yales-landemore-helps-guide-frances-unique-citizens-c onvention Custers,B.,&Fosch-Villaronga,E.(2022,July6). Law and artificial intelligence regulating AI and applying AI in legal practice.LawandArtificialIntelligence.RetrievedApril19, 2023,fromhttps://doi.org/10.1007/978-94-6265-523-2

EuropeanParliamentaryResearchService.(2019,July). Economic impacts of artificial intelligence (AI).RetrievedApril25,2023,fromhttps://www.europarl.europa.

eu/RegData/etudes/BRIE/2019/637967/EPRS_BRI(2019)637967_EN.pdf

FLI.(2017,August11). Ai Principles.FutureofLifeInstitute.RetrievedApril19,2023,from https://futureoflife.org/open-letter/ai-principles/ Franki,V.,Majnarić,D.,&Višković,A.(2023).Acomprehensivereviewofartificial intelligence(AI)companiesinthepowersector. Energies, 16(3),1077.

doi:https://doi.org/10.3390/en16031077

GlobalAssembly,TheCoreAssemblyprocess,2023,https://globalassembly.org/the-process. G,N.(2023,February16).101ArtificialIntelligenceStatistics.Techjury.RetrievedApril17, 2023,fromhttps://techjury.net/blog/ai-statistics/#gref

Harris,B.(2022,November16). Improving people's experiences through Community Forums. Meta.RetrievedApril11,2023,from

https://about.fb.com/news/2022/11/improving-peoples-experiences-through-community-fo rums/

How we work PartnershiponAI.(n.d.).Retrievedfromhttps://partnershiponai.org/ho w-we-work/

HongL,PageSE.Groupsofdiverseproblemsolverscanoutperformgroupsofhigh-ability problem solvers.ProcNatlAcadSciUSA.2004Nov16;101(46):16385-9

Ibid

ICCRulesofArbitration,January1,2021,https://iccwbo.org/dispute-resolution/disputeresolution-services/arbitration/rules-procedure/2021-arbitration-rules/#block-accordion-12

I. Digital Economy.G20ResearchGroup.(2019,June9).RetrievedApril19,2023,from

http://www.g20.utoronto.ca/2019/2019-g20-trade.html

Kimenyi,M.S.(2016,July29). Can the International Criminal Court Play Fair in Africa?

Brookings.RetrievedApril11,2023,from

https://www.brookings.edu/blog/africa-in-focus/2013/10/17/can-the-international-criminal -court-play-fair-in-africa/

36

KismetOverview.MassachusettsInstituteofTechnology.(n.d.).Sociablemachines-overview.

RetrievedApril17,2023,fromhttp://www.ai.mit.edu/projects/sociable/overview.html

Landemore,Helene.“Lecture12:ThePowerofLot.” PoliticswithoutPoliticians,Feb 23,2023. YaleUniversityLecture.

Landemore,Helene.“OpenDemocracy:ReinventingPopularRulefortheTwenty-First Century,”Oct13,2023.

Levy,S.(2022,November8).InsideMeta'soversightboard:2yearsofpushinglimits.Wired.

RetrievedApril11,2023,from

https://www.wired.com/story/inside-metas-oversight-board-two-years-of-pushing-limits/ Montréal Declaration for a Responsible Development of Artificial Intelligence.(2018). MontrealDeclarationResponsibleAI.

https://monoskop.org/images/d/d2/Montreal_Declaration_for_a_Responsible_Developme nt_of_Artificial_Intelligence_2018.pdf

NationalInstituteofStandardsandTechnology (2023). Artificial Intelligence Risk Management Framework U.S.DepartmentofCommerce.

https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf

NewYorkArbitrationConvention.(n.d.). Contracting states - list of Contracting States ContractingStates-ListofContractingStates.RetrievedApril23,2023,from https://www.newyorkconvention.org/list+of+contracting+states Nilsson,N.J.(2019).EdwardA("Ed")Feigenbaum.EdwardAFeigenbaum-A.M.Turing AwardLaureate.RetrievedApril17,2023,fromhttps://amturing.acm.org/award_ winners/feigenbaum_4167235.cfm

Nine ethical AI principles for organizations to follow.WorldEconomicForum.(2021,June23).

RetrievedApril19,2023,fromhttps://www.weforum.org/agenda/2021/06/ethicalprinciples-for-ai/ ObservatoryofPublicSectorInnovation.(2016,December17). The Irish Citizens’ Assembly. ObservatoryofPublicSectorInnovation.

https://oecd-opsi.org/innovations/the-irish-citizens-assembly/ OECD.(2019,May). AI Principles.TheOECDArtificialIntelligence(AI)Principles.Retrieved April19,2023,fromhttps://oecd.ai/en/ai-principles Ong,J.,&Drake,M.(2023,February2). EU AI policy and regulation: What to look out for in 2023.InsidePrivacy.RetrievedApril11,2023,from https://www.insideprivacy.com/artificial-intelligence/eu-ai-policy-and-regulation-what-to-l ook-out-for-in-2023/

Pause giant AI experiments: An open letter.FutureofLifeInstitute.(2023,March22).

RetrievedApril24,2023,fromhttps://futureoflife.org/open-letter/pausegiant-ai-experiments/

Pečarič,M.(2017). Can a group of people be smarter than experts?. The Theory and Practice of Legislation, 5(1), 5-29.

Providing global steerage on artificial intelligence -un.org.UnitedNationsSecretaryGeneral's RoadmapforDigitalCooperation.(n.d.).RetrievedApril24,2023,from

37

https://www.un.org/techenvoy/sites/www.un.org.techenvoy/files/general/Artificial_Intellig ence_Summary_PDF.pdf

Recommendation of the Council on Artificial Intelligence.OECDLegalInstruments.(2019,May 21).RetrievedApril19,2023,fromhttps://legalinstruments.oecd.org/en/ instruments/oecd-legal-0449

Smith,L.W.(2000).Stakeholderanalysis:apivotalpracticeofsuccessfulprojects.Paper presentedatProjectManagementInstituteAnnualSeminars&Symposium,Houston, TX.NewtownSquare,PA:ProjectManagementInstitute.

Smuha,NathalieA.,Froma'RacetoAI'toa'RacetoAIRegulation'-RegulatoryCompetition forArtificialIntelligence(November1,2019).PublishedinLaw,Innovationand Technology,Vol.13,Iss.1,2021,AvailableatSSRN:https://ssrn.com/abstract=3501410 orhttp://dx.doi.org/10.2139/ssrn.3501410

Sustensis&Czarnecki,T (n.d.). Disadvantages of Citizens’ Assembly Sustensis:Inspirationsfor Humanity'stransitiontocoexistencewithSuperintelligence.

https://sustensis.co.uk/disadvantages-of-citizens-assembly/ Taagepera,R.,Kaiser,B.,&Cervas,J.(2018).Population-DependenceofCabinetSizes. Social Science Research Network https://doi.org/10.2139/ssrn.3223745 Technica,A.(2022,December15). China bans AI-generated media without watermarks.ACM. RetrievedApril11,2023,from

https://cacm.acm.org/news/267778-china-bans-ai-generated-media-without-watermarks/fu lltext

TheEuropeanUnion(EU)GeneralDataProtectionRegulation(GDPR).HumanResearch ProtectionOffice(HRPO)|UniversityofPittsburgh.(2022,January31).RetrievedApril 11,2023,fromhttps://www.hrpo.pitt.edu/european-union-eu-general-data-protection -regulation-gdpr

UnitedNationsConventionontheRecognitionandEnforcementofForeignArbitralAwards, June10,1958,https://www.newyorkconvention.org/english Urban20.(2022,April4). 2018 - Buenos aires.Urban20.RetrievedApril19,2023,from https://www.urban20.org/u20summit/2018-buenos-aires/ Verba,Sidney,HenryE.Brady,andKayLehmanSchlozman. Unequal and Unrepresented: Political Inequality and the People's Voice in the New Gilded Age. Princeton:Princeton UniversityPress,2018.muse.jhu.edu/book/64730.

Vinuesa,R.,Azizpour,H.,Leite,I. et al. Theroleofartificialintelligenceinachievingthe SustainableDevelopmentGoals. Nat Commun 11,233(2020).

https://doi.org/10.1038/s41467-019-14108-y

WhyisAIImportant?ColoradoStateUniversityGlobal.(2021).RetrievedApril17,2023,from https://csuglobal.edu/blog/why-ai-important

38

Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.