Anthropic rejects Chinese request for access to advanced AI model Mythos

Anthropic told a Chinese think tank it couldn’t have access to Mythos, the company’s advanced AI model built for cybersecurity and software vulnerability detection. The rejection, which landed on May 11, is the latest flashpoint in a US-China technology rivalry that national security officials are now openly comparing to the Cold War nuclear arms race.

The request was informal, not a formal diplomatic overture from Beijing.

Why Mythos matters beyond Silicon Valley

Mythos isn’t your standard large language model. Its specialty is finding software vulnerabilities, the kind of exploits that can compromise everything from government networks to blockchain smart contracts.

Prediction markets on Polymarket currently show 100% confidence that Anthropic will grant Mythos access to the US government by May 31, 2026. Anthropic’s own analysis from 2025 uncovered $4.6M in blockchain smart contract exploits using Mythos-related capabilities. A model that can preemptively identify vulnerabilities in DeFi protocols is worth more than most security audits combined.

The Chinese think tank’s interest wasn’t accidental. A model that can find exploits in software can also, theoretically, weaponize them.

The crypto angle: decentralized AI gets a tailwind

White House proposals floated on May 4 for AI vetting frameworks could indirectly boost decentralized AI projects. If centralized AI companies face increasing pressure to restrict access based on nationality, projects building open, permissionless AI infrastructure become more attractive to developers and nations locked out of the US ecosystem.

Anthropic’s tokenization problem

On May 12, Anthropic declared unauthorized stock tokens purporting to represent Anthropic equity as invalid. These were unauthorized tokens that mimicked Anthropic’s equity, creating synthetic exposure to a private company without its consent. Anthropic warned of potential fraud associated with these tokens.

What this means for investors

If Anthropic’s 2025 analysis found $4.6M in smart contract exploits, the question becomes who else is building similar tools, and whether DeFi protocols will start integrating AI-powered vulnerability scanning as standard practice.

The unauthorized stock token incident should give pause to anyone buying tokenized exposure to private AI companies. If the issuing company itself calls the tokens fraudulent, liquidity can evaporate faster than you can close a position. Regulatory clarity around tokenized equity remains thin, and Anthropic’s response suggests that major AI firms won’t quietly tolerate unauthorized financialization of their brands.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.