OpenAI has begun granting a select group of users access to a new AI model designed to be more effective at identifying software security vulnerabilities, coming just a week after rival Anthropic PBC announced a limited rollout of its own AI system, Mythos.

The ChatGPT maker said on Tuesday that it is starting to deploy GPT-5.4-Cyber, a model focused on detecting flaws in software to help organisations address them more quickly. OpenAI noted that GPT-5.4-Cyber places fewer restrictions on how users can interact with it for security testing purposes.

The model will be made available to participants in OpenAI’s Trusted Access for Cyber programme, launched in February to give select customers and cybersecurity professionals early access to its most advanced tools.

OpenAI plans to expand the number of participants in its early access programme, initially allowing hundreds of users to test the new model before scaling up to thousands in the coming weeks.

OpenAI and Anthropic are competing to build increasingly advanced AI systems capable of a broad range of tasks, while also seeking to attract enterprise customers willing to pay for access.

However, as their models improve at coding and identifying vulnerabilities in software, concerns have grown about potential misuse by criminals and state-backed hackers.

AI tools are already being leveraged in parts of the cyberattack lifecycle, raising further security concerns among researchers and policymakers.

Last week, Anthropic introduced an AI model called Mythos (Mythos), which it says specializes in identifying and exploiting vulnerabilities across operating systems and web browsers.

The company has limited access to a small group of trusted partners, including Amazon.com Inc., Apple Inc., and Microsoft Corp., with the aim of enabling them to use the model to detect security weaknesses and strengthen their systems.

Mythos has quickly raised concerns among financial institutions and government agencies.

During a meeting last week with Wall Street executives—convened by U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell—industry leaders were cautioned that the Mythos model should be taken seriously.