The UN Security Council usually talks about wars, ceasefires and sanctions
On 23 September it heard from the people who build AI.
France holds the Council presidency this month, and its foreign minister, Jean-Noël Barrot, chaired the session. The briefers were Sam Altman of OpenAI, Dario Amodei of Anthropic (by video link), Clément Delangue of Hugging Face, and the AI scientist Yoshua Bengio. Reuters reported that China's DeepSeek and Moonshot were also invited to give statements.
Eleven days earlier, Amodei had published an essay asking the industry to slow the pace at which AI becomes more capable. Altman agreed. The people building the car are asking for brakes.
Bengio's proposal to the Council: license frontier AI the way we license medicine, aviation and nuclear power, with compulsory liability insurance and reporting of incidents.
Reuters reported that China's DeepSeek and Moonshot were also invited to give statements.
My first thought was that my own profession already works under rules like that. In the UK, to call myself an architect I have to be registered and carry professional indemnity insurance. Since the Building Safety Act, serious safety occurrences on higher-risk buildings have to be reported to the regulator. We ask that of a block of flats. We don't yet ask it of systems used by hundreds of millions of people.
If I had a seat at that table, I'd ask three things:
- When it goes wrong, who pays?
- Who can stop it: the company, or someone outside it?
- You say slow down. Slow to what, and who measures it?
What would you have asked them? Tell me what you think in the comments 👇
