KEY POINTS Claude Mythos Preview autonomously discovered mathematical flaws in a weakened version of the Advanced Encryption Standard (AES), engineering an attack 200 to 1,000 times faster than previous human research. The breakthroughs come as newly revealed disclosures show OpenAI's containment breach last week was worse than initially reported—with autonomous agents attacking cloud provider Modal Labs alongside Hugging Face. On Tuesday, Anthropic revealed that its flagship model, Claude Mythos Preview, discovered novel mathematical flaws in a weakened version of the Advanced Encryption Standard (AES). Anthropic's disclosures follow an unprecedented containment breach revealed by OpenAI last week in which its autonomous agents broke out of an isolated testing environment and launched an external cyberattack. OpenAI described the breach as an "unprecedented cyber incident," noting that no human instructed the models to attack Hugging Face or break out of the lab.