On September 22nd, 2024, Atlas grassroots consultations to hear from the people of New York and the world what they want the United Nations to do when it comes to regulating artificial intelligence. Through a three-hour-long event, we learnt, debated, and agreed on a set of recommendations democratically. Around 65 people attended the event, both online and in person.
Opening
Our co-founder Colombe Cahen-Savaldor opened the day by talking about the global threats AI creates, the urgency of regulating AI globally and the need for the United Nations to take the leadership on this. She discussed the current approach taken by the UN, its flaws and the lack of citizens’ representation, and how the current system could be improved. She then went through the programme of the day, the “rules of the game” and explained how the results of the consultations would be used to inform Atlas’ positions on the issue.
Panel
The event was opened with thought-provoking remarks by the following speakers: Anna Christmann, Member of the German Parliament, Coordinator for aerospace in the federal government, member of the United Nations High-Level Advisory Body on Artificial Intelligence, Santiago Garces, CIO of the City of Boston, Senior Fellow at the Burnes Center for Social Change and Technical Lead for AI for Impact, Fellow at Harvard Kennedy School, Christine Keung, General Partner at J2 Ventures, Aspen Institute Fellow, former Chief Data Officer for the City of San Jose, CA. The panel was moderated by Kyle Johnson, Director of the International Strategy Forum.
The panel started with a discussion on the main issues facing AI, from its definition to the first attempts to regulate it, looking at the EU. Anna Christmann mentioned the report “Governing AI for humanity” just published by the United Nations High-Level Advisory Body on Artificial Intelligence as a set of guidelines the world should follow. Santiago Garces focused on the role local governments could play in using AI to improve services to citizens. Christine Keung brought in a private sector perspective, discussing the need for the market to have clear regulatory norms on the use of data used by AI. Several specific policy solutions were also aired during the conversation.
Such panels can help in setting the stage and ensuring that participants think outside of the box.
Identification of Issues
Following this panel, participants were asked to identify at least three issues around AI. People were divided into two groups and asked to designate a spokesperson to take notes and present the results.
The spokesperson of each group was then asked to present the issues identified. Here are the main issues/pain points that came up:
- Replacement of Jobs and difficulties in re-skilling people
- Lack of trust and transparency in governments/the UN in regulating AI, and companies using it
- Lack of enforcement/accountability mechanisms to ensure humanity’s safety when it comes to AI
- Military usage of AI and surveillance
- Lack of easy way to identify what is AI generated
- Misinformation and disinformation
- “Doomsday” scenarios and survival risks for humanity
- Lack of equity for AI to benefit the many
Trust, enforcement and equity were prioritised.
Creation of Citizens’ Proposals
Following the prioritization of issues, citizens went back to their group to create proposals that would best address any or all of the three issues selected. They were asked to make concrete proposals.
They were then asked to present their solutions to the audience again. Here are the solutions that came up:
- Overall, there seemed to be an agreement that regulations must come from the global level, through an AI treaty or similar. We discussed that simply agreeing on a broad set of principles was not enough, and that adequate regulations must be adopted for humanity to stand a chance. People seemed to agree that a global citizens’ assembly would help in setting key principles to be included in the AI treaty.
-
On trust and transparency: we need to create a culture of honesty, trust and transparency. For this, any future UN treaty must:
- Have strong whistleblower protection for companies and government, to ensure the public is made aware of any public risk.
- Include public audits, that could be collaborative and AI-powered for citizens to spot problems in the use of AI by multinational companies or governments.
- Create rules on how data can be used by companies to not only protect citizens’ interest, but create enforcement mechanisms (see below)
-
On enforcement and accountability: it is fundamental for those rules to be enforceable, not just a set of principles. For this, any future UN treaty must:
- Include liability regimes: those could be individual criminal liability and/or company liability.
-
On this point, we did not have time to agree, but there was a debate on which approach to adopt:
- A. tracking AI chips/compute systems and deciding who gets them
- B. blocking datasets to some companies not respecting the AI treaty. This would create incentives to comply to have access to one of the most valuable assets: data.
-
On equity:
- The data protection point above was particularly relevant when talking about African countries who would benefit from having higher data protection standards.
- The citizens’ assembly point was picked up again to ensure adequate representation in deciding AI rules and principles.
Next steps
- These solutions will now be given to Atlas’ policy team to refine/review, and our membership will then vote on them at our next Summit.
- We will bring those demands to the UN, national governments, and share them widely!