OpenAI and Anthropic Launch High-Capability AI Cyber Models
OpenAI and Anthropic have released specialized AI models capable of discovering software vulnerabilities, prompting urgent security reviews by global financial regulators and government agencies.
AI laboratories Anthropic and OpenAI have launched specialized models designed to identify and exploit software vulnerabilities, sparking concerns over a new AI-enabled cybersecurity arms race. Anthropic released Claude Mythos Preview, a model capable of discovering thousands of high-severity zero-day vulnerabilities, including a 27-year-old Linux kernel flaw. To mitigate risks, Anthropic launched Project Glasswing, granting restricted access to approximately 50 vetted partners, including Microsoft, Google, and JPMorgan Chase.
In response, OpenAI released GPT-5.4-Cyber, a cyber-permissive model designed for defensive tasks and binary code reverse engineering. While Anthropic maintains a tight circle of access, OpenAI is pursuing a broader democratization strategy through its Trusted Access for Cyber program, providing the tool to thousands of verified security professionals and partners like CrowdStrike and NVIDIA.
The capabilities of these models have alarmed global officials. In the United States, Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell summoned Wall Street executives to ensure legacy banking systems are defended. Similar risk assessments were conducted by the European Central Bank, the Bank of England, and the Bank of Canada. While the National Security Agency has begun using Mythos, the Pentagon issued a formal supply-chain risk designation against Anthropic. OpenAI has since begun briefing federal agencies and Five Eyes intelligence partners on the deployment of its defensive tool.
