Anthropic says its Mythos model found vulnerabilities in cryptographic algorithms that secure the internet
Anthropic's Claude Mythos Preview identified vulnerabilities in cryptographic algorithms, including a new attack on the HAWK post-quantum scheme. The model completed the analysis in 60 hours for approximately $100,000 in API costs.
Anthropic revealed that its Claude Mythos Preview model successfully identified weaknesses in cryptographic standards. Specifically, the AI discovered an improved attack vector against HAWK, a post-quantum digital signature scheme previously scrutinized by human experts for over two years.
This development highlights the growing capability of large language models to perform specialized technical tasks traditionally reserved for human cryptographers. The model achieved this result in roughly 60 hours, significantly faster than typical human review cycles.
The analysis incurred an estimated API cost of $100,000, suggesting that while AI can accelerate security research, it remains a resource-intensive process. These findings could influence how organizations approach post-quantum cryptography and AI-driven security auditing in the future.
This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.