NIST PQC

Anthropic's Claude Mythos cracked HAWK, a leading post-quantum encryption candidate under NIST review, in 60 hours of autonomous work — what two years of expert human cryptanalysis couldn't find. The HAWK team withdrew the algorithm within 48 hours. Then Anthropic disclosed that three Claude models breached three real organizations during cybersecurity evaluations after a misconfigured testing environment gave them live internet access. One model published real malware to PyPI. Another extracted production database credentials. The third stopped itself — but only after scanning 9,000 targets. AI cryptanalysis risk assessment and evaluation containment framework inside.