ARC & International Legal Consultant

Delineating stronger parameters around the use of Ai in conflict (semi- autonomous and autonomous weapons), improved protocols for verifying information online including considering delays before publication, and finding a balance between free speech protections and heightened protections from Ai harms. A successful first Global Dialogue would 1) provide language that can be proposed for domestic legislation that provides enough potential to harmonize law across jurisdictions while addressing context specific protections, 2) offer guidance on ways to operationalize these protections, and 3) suggest improved monitoring and evaluation mechanisms.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

Safe, secure and trustworthy Ai is important because of the rise in harms to children which could partly be remedied by better age gates and access verification tools. The rise of Ai bots for uses such as friends, romantic partners, and psychological/counseling providers combined with the tendency of hallucinations has increased mental health risks to users. Social, economic, ethical, linguistic and technical implications is important because 1) with the expanding uses of Ai, there are a greater need for equal access to Ai technology for remaining engaged and competitive in society as those without access may fall behind; 2) there is a need for LLM's exposure to many different cultures and ethnicities to reduce the risk of bias in the data provided to users; and 3) a wider range of languages used in Ai application creations means that it will be more accessible to a wider populations. Protection and promotion of human rights is important because imbedding the UNGPs and the human rights language of global frameworks including Ai laws in France, the US and Japan means that there will be incentives to uphold these rights for the greater populations from a rights-based approach while providing avenues for accountability from a responsibility-based approach. Finally, transparency is important from a legitimacy perspective in ensuring that people understand who is responsible for creating the legal framework, the motivations for and goals of the frameworks. Accountability mechanisms, as part grievance and remedy mechanisms are important for ensuring that the laws will be adhered to and that when they aren't someone or some entitty will be held responsible. Human oversight is important for responsibility and accountability purposes, but also because the nature and context of the Ai's use would determine the degree of its autonomous use.

No, this list is sufficiently comprehensive because there are cross-cutting and emerging issues under most of these listed.

In the US, Ai regulation is largely a state-by-state patchwork which means there are no uniform ways to safeguard access to Ai tools, regulate around the uses of Ai for mental health, medicine, contracts and many other areas of life. In law, we are still looking at how to address Ai's uses for drafting awards, its uses in arbitration and in ADR practices, the use of Ai for parts of trials, sentencing recommendations and many other aspects of the legal profession.

• In advancing international cooperation on Ai governance, Ai translations to reach the broadest sets of global commnities can exponentially increase inclusivity
• it can be used to generate insights into ways to address the most challenging harms by region
• it can more accurately assess and predict many types of risks
• it can help standardize monitoring such that even if there are variants at the most local levels, Ai can derive a process for standardization at national and regional levels.

The Stargate partnership infrastructure project between OpenAi, Softbank, Oracle and Microsoft- supporting the partnership but also addressing the environmental impacts such as the overuse of water by data centers and the environmental degradation caused by building more centers (loss of green spaces). Supporting Ai partnerships between tech and medical institutions for improved surgeries, diagnosis, treatment plans and research. Building on the Ai's uses for national and international security, especially in regards to trade, manufacturing and intelligence gathering with improved guardrails.

Some of the best ways for different stakeholders to contribute to the dialogues is to 1) do a call for inputs to permit stakeholders to share their concerns and perspectives similar to this form along with their contact information so that it can be included in the discussions, 2) provide a virtual option for attendance for those who can not travel, and 3) although requiring more expenditures, hold regional Ai dialogues that may increase accessibility.

Among the many underrepresented communities in Ai governance discussions are various groups in Global South generally with some of that caused by inconsistent energy access in some places, LGBTQ+ members, low-income society members, certain racial minorities even though this marginalization is more often connected to their socio-economic status within those minority groups, indigenous groups, and people with disabilities.

Innovative formats that would not only foster but increase meaningful and dynamic dialogue engagement would include virtual accessibility, multiple breakout discussions with outcome goals as a work product, fish bowl discussions with some debate style elements, a session in which there is a problem to which teams present/pitch possible solutions (somewhat shark tank style), and live polls during session presentations.

Some good policies would include those with 1) accountability structures displaying named owners of Ai systems, with clear definitions of who deploys the Ai and grant exceptions, pause and retire systems which partly overlaps with human oversight; 2) governance frameworks including specific accountability and remediation provisions to address non-reversibility of actions, open-ended decision-making pathways and privacy vulnerabilities; 3) a deeply comprehensive compliance framework including mapping controls adhering to applicable standards such as those set out in the EU Ai Act and US sector specific laws; and 4) a cross-functional governance committee comprising Risk, Legal, IT and Security, data science and key business units which will all convene and share information and work together in monitoring, evaluation and implementation of improvements.