Most crypto firms can't get the good AI, and the security math is getting scary 🔐

Most of crypto's largest firms have yet to obtain access to the most capable frontier AI models, leaving only a handful of companies with the tools to audit their own code, Binance's chief security officer Jimmy Su told Cointelegraph. Coinbase said in June that it had secured access to Anthropic's restricted Mythos model, and Zcash's Zooko Wilcox said Anthropic used the model to audit the Zcash protocol at the request of Shielded Labs. Other major crypto players, by contrast, remain on the outside. "That's one advanced frontier model that hasn't been made available to crypto just yet," Su said. "We have been trying to make inroads there. We also talked to other crypto exchanges and our own investors to try to make some progress. But we haven't gotten the most frontier AI model, like Mythos."

The uneven access is creating a new security gap in an industry where exploits can put billions of dollars at risk. Anthropic and OpenAI have both restricted their most cyber-capable models from the public, while increasingly capable open-source alternatives raise concerns that crypto firms may face sophisticated AI-assisted attacks without equivalent defensive tools. Anthropic says Mythos 5 uses the same underlying model as its publicly available Fable 5 but without safeguards that restrict sensitive cybersecurity work. OpenAI operates a similar tiered system: verified defenders can use GPT-5.5 with "Trusted Access for Cyber," while its more permissive GPT-5.5-Cyber model is reserved for a smaller group conducting authorized penetration testing.

Crypto security executives interviewed by Cointelegraph said restricting frontier cyber models is likely necessary at first, but becomes harder to justify once publicly available models approach the same capabilities. Su said Anthropic's controlled rollout is a responsible approach because newly released models may benefit attackers faster than defenders. "If it enhances the attacker much faster than the defender, then it actually is harming the ecosystem," he said, adding that a limited testing period could reduce the potential "blast radius." He added that the calculation changes when competing models become more powerful and widely available. "As other more powerful models are being released, the pressure will be on Anthropic to make it more widely available," Su said.

The stakes are already visible in vulnerability data. The number of critical-severity CVEs has climbed after the launch of Claude Mythos Preview, according to Epoch AI. Solana Foundation chief information security officer Michael Coates, who joined the foundation in July, also supported safeguarded access as a near-term measure, though the broader question of how long such restrictions can hold remains open across the industry.