The crypto industry is competing for Mythos test access and using existing AI to detect system vulnerabilities.

According to 1M AI News monitoring, Anthropic has opened early testing of Mythos to tech and financial giants such as Amazon, Apple, Google, and JPMorgan, but major crypto trading platforms and custodians have not been granted access. Several crypto companies are striving to gain access, while also using existing AI models to conduct security checks on their systems.

Philip Martin, Chief Security Officer of an exchange, said the company is in "close communication" with Anthropic regarding Mythos, and noted that "Mythos and similar future models will enable deeper and more scalable software and system tests, which will accelerate both digital threats and defenses." Jimmy Su, Chief Security Officer of another exchange, stated that the company is testing system security from both an external attacker's perspective and at the source code level, using both commercial models, including Anthropic, and their own internally developed models.

Crypto custody technology firm Fireblocks (which provides custody infrastructure for clients such as BNY Mellon) has used Claude Opus 4.6 for vulnerability detection. CEO and co-founder Michael Shaulov said this model identified problems that previous security audit firms had missed, and all vulnerabilities have now been fixed. He expects Anthropic will open access to the crypto industry before a broad release of Mythos. Decentralized exchange Uniswap, which holds $3.3 billion in assets, is also trying to contact Anthropic. CEO Hayden Adams publicly posted for help last Friday on X: "Who has a good contact at Anthropic?"

Goldman Sachs CEO David Solomon proactively mentioned Mythos on Monday during the earnings call: "We are aware of Mythos and its capabilities, we have obtained the model, and are working closely with Anthropic and all security vendors." While banks are already testing, the crypto industry is still waiting in line, reflecting Anthropic's cautious approach to distributing Mythos and also highlighting the crypto infrastructure's sense of urgency regarding AI security threats. Niv Hoffman, co-founder and CTO of Sequoia-backed cybersecurity startup Buzz, pointed out that with much of the crypto industry's infrastructure being open-source, the attack surface is naturally larger. The vulnerability exploitation capability of AI models means that attackers could complete the entire vulnerability discovery-to-exploitation process within hours or even minutes in the future.