Anthropic has announced that its experimental AI model, Claude Mythos Preview, identified a previously unknown cryptographic weakness in HAWK-256, a post-quantum digital signature scheme, after approximately 60 hours of analysis. The discovery has reignited debate over the future of cryptographic security and whether blockchain networks such as Bitcoin should accelerate preparations for an era shaped by increasingly capable AI systems.

According to the company, the AI model operated with only occasional human guidance and identified an attack that effectively halves HAWK-256's security margin. Anthropic estimated that the research consumed roughly $100,000 worth of API compute resources.

Key Findings

The research produced several notable results:

• A new key recovery attack against HAWK-256 that substantially lowers its effective security level.
• A 200 to 800 times faster attack on a seven-round version of AES-128 — not the full 10-round AES-128 standard used in production.
• Advance disclosure of the findings to the algorithm's designers as well as relevant government and industry organizations prior to publication.

Researchers emphasized that HAWK remains a proposed post-quantum signature scheme rather than a widely deployed standard. Post-quantum cryptography is an active field driven by concern that future quantum computers could break widely used public-key systems such as RSA and elliptic curve cryptography. The U.S. National Institute of Standards and Technology has been running a multi-year post-quantum standardization process since 2016 and published its first finalized post-quantum standards in August 2024. HAWK is among numerous proposals under academic study but has not been selected as a NIST standard.

Similarly, the AES result applies only to a deliberately weakened research variant, not the encryption currently safeguarding banking systems, messaging platforms, and internet traffic. Cryptanalytic advances against reduced-round versions of AES are a common academic technique for probing the algorithm's structure and do not imply vulnerabilities in the full standard.

Implications for Bitcoin and Blockchain Security

The findings prompted fresh discussion within the cryptocurrency community, given that Bitcoin's long-term security depends on robust cryptography. Some observers argue that increasingly capable AI systems could shorten the timeline for discovering vulnerabilities in cryptographic algorithms, making proactive upgrades more critical.

However, the new discoveries do not threaten Bitcoin's current cryptography. Bitcoin relies primarily on SHA-256 hashing and the secp256k1 elliptic curve signature system, neither of which was affected by Anthropic's research. The HAWK attack targets an unrelated post-quantum signature proposal, while the AES result concerns a reduced-round academic variant rather than any deployed encryption standard.

The work nonetheless highlights how frontier AI models are emerging as powerful research tools — capable of both uncovering vulnerabilities and contributing to stronger future cryptographic standards. The ability to conduct approximately 60 hours of sustained cryptanalysis for a reported $100,000 in compute also signals a potential shift in the economics of security research, where tasks once requiring weeks of expert human effort may increasingly be augmented or accelerated by AI systems. As AI capabilities continue to advance, cryptographers and the cryptocurrency industry may face growing pressure to evaluate and adopt more resilient security mechanisms before theoretical risks become practical ones.