"Trust us" isn't a safety framework.

A handful of labs control the models, the compute they train on, and the benchmarks they grade themselves against.

That's not accountability. And it won't lead to greater safety.

Real accountability requires transparency, openness, and access.

That's exactly what we're building at @ionet.
https://io.net/blog/ai-safety-is-not-the-same-as-ai-accountability