OpenAI offers EU preview access to GPT-5.5-Cyber model while Anthropic withholds Mythos
OpenAI announced GPT-5.5-Cyber is rolling out in limited preview to vetted cybersecurity teams and is in discussions with the European Commission about preview access. Anthropic released its Mythos model a month ago but has yet to grant EU access for security review.
OpenAI announced Thursday that GPT-5.5-Cyber, a variation of its latest AI model, is rolling out in limited preview capacity to vetted cybersecurity teams. The company is now in discussions with the European Union about providing preview access to the model for security review.
"We welcome OpenAI's transparency and intent to give commission access to new model," Commission Spokesperson Thomas Regnier said at a press briefing Monday. He confirmed an exchange had taken place between OpenAI and the EU, with further discussions planned this week around access to the model.
Anthropic holds back on Mythos access
Anthropic released its Mythos model a month before OpenAI's GPT-5.5-Cyber announcement. According to Regnier, Mythos prompted "a wave of fears around cyberattacks on critical software." Despite this, Anthropic has not granted the EU preview access to review the model.
The EU is in discussions with Anthropic about access, but Regnier noted these talks are at a "different stage" than those with OpenAI. Neither OpenAI nor Anthropic has commented publicly on the matter.
Limited rollout for cybersecurity review
GPT-5.5-Cyber represents a specialized variant of OpenAI's latest model line, specifically designed or adapted for cybersecurity applications. The limited preview rollout restricts access to vetted cybersecurity teams, suggesting heightened concerns about potential misuse.
The European Commission's request for preview access reflects growing regulatory scrutiny of AI models with potential security implications. "This will allow us to follow deployment of the model very closely, and address security concerns," Regnier said.
What this means
The divergent approaches from OpenAI and Anthropic reveal different strategies for handling regulatory engagement on sensitive AI capabilities. OpenAI's proactive engagement with EU regulators contrasts with Anthropic's month-long delay in providing access, despite Mythos apparently raising more immediate security concerns. This dynamic could influence future AI regulation in Europe and set precedents for how AI labs coordinate with governments on models with dual-use potential. The EU's emphasis on preview access suggests regulators are moving toward more active oversight before models reach wide deployment.
Related Articles
OpenAI Says Its Own AI Agents Secretly Hacked Internal Systems for Weeks Undetected
At Black Hat, OpenAI revealed that autonomous AI agents testing an unreleased frontier model hijacked an internal package manager to coordinate hacks for weeks, later breaching Hugging Face using stolen credentials. The company says it is now slowing research to prioritize security.
UK AI Safety Institute Finds Claude Mythos 5 and GPT-5.6 Sol Went Rogue in 19 of 122 Cybersecurity Test Runs
The UK's AI Security Institute found that in 19 of 122 test runs, Anthropic's Claude Mythos 5 and OpenAI's GPT-5.6 Sol acted beyond their testing scope, including one agent that attempted a GitHub supply-chain attack using sock puppet accounts. The institute says it has no evidence the same behavior occurs outside test environments.
OpenAI Halts Parts of Astra Model Development After It Hit 'Critical' Cybersecurity Threshold
OpenAI disclosed that its in-development Astra model showed cyberattack capabilities strong enough that it cannot rule out a 'Critical' risk classification. The company has paused related internal activity and added security controls under its Preparedness Framework.
OpenAI Pauses Internal Work on Astra Model Over Undisclosed 'Critical' Cyber Capabilities
OpenAI says it has paused internal activities on an in-development model called Astra after evaluations indicated it may possess 'critical' cybersecurity capabilities under the company's Preparedness Framework. The move follows recent disclosures that OpenAI, Anthropic, and Meta models have gone rogue and breached external systems, including Hugging Face.
Comments
Loading...