Bottom line: Claude Mythos Preview has achieved a practical end-to-end attack against the post-quantum scheme HAWK-256 and massive accelerations in AES attacks.
Anthropic announces that Claude Mythos Preview has derived a complete key recovery attack against the post-quantum scheme HAWK-256 as well as an acceleration of the 7-round AES-128 attack by a factor of 200 to 800. The HAWK attack exploits a previously unexploited symmetry in the lattice structure underlying the signature scheme.
Anthropic has developed a complete key recovery attack against the post-quantum scheme HAWK-256 with Claude Mythos Preview. The attack exploits a previously unexploited symmetry in the lattice structure underlying the signature scheme. The reference implementation provided by Anthropic executes the attack on a 96-core server with an expected runtime of approximately three hours and 42 minutes.
In parallel, Claude achieved massive acceleration of existing attacks on 7-round AES-128: the procedures are between 200 and 800 times faster than previous approaches. This demonstrates that generative AI models are capable of identifying and exploiting cryptographic weaknesses that might have been overlooked by conventional analysis.
For CTOs and security officers, this development has far-reaching implications: the ability of AI models to automatically generate attack vectors against established cryptographic schemes underscores the need for enhanced evaluation of post-quantum schemes prior to their standardization and broad deployment. HAWK-256 was intended as a candidate for post-quantum cryptography; this breakthrough requires a reassessment of such approaches in production environments.
Source: thehackernews.com · Published 28 July 2026
Lumi AI News — AI-assisted curation pursuant to Article 50 EU AI Act. Paraphrasing and classification by Lumi News Pipeline v1.7.3.