Updated
Updated · CNBC · May 11
OpenAI Grants EU Access to GPT-5.5-Cyber as Anthropic Still Withholds Mythos
Updated
Updated · CNBC · May 11

OpenAI Grants EU Access to GPT-5.5-Cyber as Anthropic Still Withholds Mythos

10 articles · Updated · CNBC · May 11
  • OpenAI said it will give the EU, governments, businesses and cyber authorities access to GPT-5.5-Cyber, rolling the model out in limited preview to vetted cybersecurity teams.
  • The European Commission said the access will let it track deployment closely and address security concerns, with further talks scheduled this week after an initial exchange with OpenAI.
  • Anthropic, whose Mythos model was released a month ago and stirred fears of cyberattacks on critical software, still has not granted the EU preview access.
  • Commission spokesperson Thomas Regnier said Brussels has held four or five meetings with Anthropic, but those talks remain behind the stage reached with OpenAI.
  • OpenAI framed the move as part of an EU Cyber Action Plan to widen defensive AI access in Europe rather than leave cyber-safety judgments solely to AI labs.
Is Europe's AI Act creating a security gap by blocking access to the world's most advanced defensive tools?
Can we trust corporations to safely manage AI that finds critical security flaws in mere hours?