Ahead of Sam Altman's UN address, OpenAI proposes new ways to track AI misalignment risks - Axios

Oh my goodness, buckle up, because OpenAI has just *written the rules* — before anyone even asked it to! ✨ On Monday the company released a set of international AI safety standards, and rather than guarding this treasure, it is recommending that AI safety institutes around the entire world be brought in to help with standard-setting. That's not a company marking its own homework; that's a company handing out pencils to the whole classroom! And the chosen vehicle? The U.S. Center for AI Security and Innovation, part of the Commerce Department — safety and commerce, now *roommates*, and honestly, who understands the value of a thing better than the people who price it?

There's more! Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng have been chatting about a U.S.-China AI dialogue, and this one comes with the most thoughtful feature imaginable: each country notifies the other when an AI incident reaches national security risk levels. That is pen-pal diplomacy at the absolute highest level! Some people will ask 'why only notify once it's already a national security risk level,' and to them we say: do you not appreciate a *confirmed* signal? Two of the world's biggest AI powers, practising the beautiful art of telling each other things — warmly, and slightly after the fact. ✨

And the whole delightful package arrives just ahead of Sam Altman's address to the United Nations, which means the standards get to travel *business class* to the biggest stage on Earth! Some might note that these standards come from a company that would, under most futures, be on the receiving end of standards — but that's just vertical integration of virtue! Who better to advise the safety institutes than the company with the most to be safe about? The Commerce Department centre has a beautifully long name, the rules are international, voluntary and *adorable*, and the notification system means we will definitely all hear about it eventually. What a time to be standardised. ✨