PANews reported on May 6th that, according to Decrypt, Anthropic CEO Dario Amodei warned in a conversation with JPMorgan Chase CEO Jamie Dimon that AI is discovering software vulnerabilities faster than organizations can patch them, potentially creating a dangerous window of six to 12 months. Anthropic's Mythos model has identified tens of thousands of software vulnerabilities, with an early version finding 271 vulnerabilities in the Firefox browser in a single scan. Amodei stated that the danger lies in the potential for a significant increase in the number of vulnerabilities, the number of data breaches, and the financial losses caused by ransomware.

Mythos can simulate multi-step cyberattacks without human intervention in controlled testing. Anthropic has limited its model to a small group of partners and is patching vulnerabilities through Project Glasswing before similar tools become widely available. OpenAI CEO Sam Altman believes concerns may be exaggerated, calling Anthropic's use of "fear-based marketing" to justify restricting access to the technology.