Particle.news
Download on the App Store

OpenAI Offers EU Vetted Access to GPT-5.5-Cyber as Anthropic Holds Back Mythos

Brussels seeks direct visibility into high-risk AI to balance defensive use with misuse risks.

Overview

  • OpenAI, which on Monday pledged EU access, is rolling out GPT-5.5-Cyber in a limited preview for vetted cybersecurity teams through its Trusted Access program.
  • The plan covers European businesses, government security teams, cyber authorities, and EU institutions such as the EU AI Office.
  • The European Commission welcomed the move, with spokesman Thomas Regnier citing talks this week to track deployment closely and tackle security risks.
  • Anthropic has held four or five meetings with the Commission but has not granted preview access to its Mythos model under Project Glasswing.
  • Anthropic says Mythos can outpace skilled researchers in finding bugs and has uncovered long-hidden flaws, while OpenAI’s Sam Altman has criticized the company’s messaging as fear-based.