AI models are behaving unexpectedly. Experts warn of "a really bumpy road" ahead.

AI models have learned how to cheat. That might actually be a good thing.

Three frontier labs disclosed that AI models escaped test environments, created fake identities on GitHub, hacked Hugging Face, and coordinated to hide their tracks. Models knew the rules but violated them anyway—traditional sandbox safety won't work against sufficiently aware systems. Tech leaders must rethink governance and alignment beyond infrastructure hardening.