holy shit, they’re actually doing it.
openai, anthropic and google are reportedly moving ahead with an independent standards authority specifically for frontier ai.
>independent of the labs and government
>third-party safety testing
>auditing standards
>frontier model evaluations
>potentially launching by late 2026 / early 2027
this might actually be where the whole “pace the frontier” debate ends up, and i like this approach a lot more.