OpenAI Fires Back at Anthropic With Its Own Cybersecurity AI — Here’s What We Know

TLDR

• OpenAI released GPT-5.4-Cyber to a select group of verified cybersecurity professionals
• The launch comes one week after rival Anthropic released its own security-focused AI model, Mythos
• Anthropic’s Mythos raised concerns with U.S. Treasury Secretary Scott Bessent and Fed Chair Jerome Powell
• OpenAI plans to expand access from hundreds to thousands of verified defenders in coming weeks
• Nearly one in three U.S. businesses paid for Anthropic’s tools in March, per Ramp data

OpenAI has released a new AI model built specifically for cybersecurity work. The model, called GPT-5.4-Cyber, is now available to a small group of verified users.

The release came on Tuesday, April 15. It follows Anthropic’s limited launch of its own security model, Mythos, by just one week.

GPT-5.4-Cyber is described as a “more permissive” version of GPT-5.4. It is designed to lower the refusal boundary for cybersecurity tasks that OpenAI considers legitimate defensive work.

Access is currently limited to participants in OpenAI’s Trusted Access for Cyber program. That program launched in February and requires identity verification for users who want to work with the model.

OpenAI said it plans to grow the program from hundreds of testers to thousands of verified defenders in the coming weeks. The company said it will continue to require identity checks before granting access.

OpenAI began adding cyber-specific safeguards to its models in 2025. In March, it launched Codex Security to help identify and fix software vulnerabilities.

Concerns Over Anthropic’s Mythos

Anthropic’s Mythos model has already drawn attention from top U.S. officials. Last week, Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell warned Wall Street leaders to take the risks from Mythos seriously, according to Bloomberg.

Anthropic chose to offer Mythos to around 40 large businesses to help improve their cyber defenses. The company limited access due to concerns about the model falling into the wrong hands.

Ireland’s National Cyber Security Centre director told a parliamentary committee on April 14 that models with similar capabilities to Mythos should be expected from bad actors before the end of the year.

Anthropic co-founder Jack Clark said other companies will release similar systems within months. He added that open-weight models from China with these capabilities could follow within a year to 18 months.

OpenAI’s Enterprise Push

OpenAI has been focusing more on enterprise customers, a market where Anthropic has gained ground quickly. Data from payments company Ramp shows nearly one in three U.S. businesses paid for Anthropic tools in March.

OpenAI has been cutting back on less profitable projects. The company has paused plans for an adult version of ChatGPT and put its Stargate UK project on hold.

Microsoft, OpenAI’s largest backer, has agreed to rent data center capacity at a site originally planned for the Stargate Norway project. A separate deal with UK AI infrastructure firm Nscale also fell through.

OpenAI announced it will open its first permanent London office in 2027, with space for more than 500 people. The company said London will become its largest research hub outside the United States.

OpenAI has also announced a new “superapp” inspired by Anthropic, and launched dedicated AI health tools shortly after Anthropic released its own health-focused Claude features.

The post OpenAI Fires Back at Anthropic With Its Own Cybersecurity AI — Here’s What We Know appeared first on CoinCentral.