In brief: OpenAI is offering the EU access to an advanced cybersecurity AI model, while Anthropic’s more restrictive access policy for its Claude model in Brussels is causing frustration.
ChatGPT-maker OpenAI is in discussions with the European Commission to grant EU authorities access to a model capable of identifying software security vulnerabilities. This comes at a timely moment for Europe, as cybersecurity regulation struggles with new AI risks.
OpenAI’s head of initiatives, former British Chancellor George Osborne, wrote to the Commission between Sunday and Monday to offer access to the advanced AI model capable of exploiting software security vulnerabilities. The company has also already begun contacting member states.
The initiative is intended to help Europe better protect critical infrastructure, public institutions and utilities. The move eases concerns in Brussels, where cybersecurity and AI officials have had no access to Anthropic’s Claude model for weeks.
A European Commission spokesperson welcomed OpenAI’s “transparency” and emphasized that this would help EU officials monitor the deployment of the model and address potential security concerns.
The offer is based on a cybersecurity-focused variant of GPT-5.5, OpenAI’s most advanced model, which was introduced about three weeks ago. EU officials will receive access to the least restricted version of the model, while the main model contains extensive safeguards against misuse.
OpenAI’s move contrasts sharply with Anthropic, which has made its Claude model available to only a handful of US companies and forty unnamed organizations. This has frustrated EU officials and lawmakers, who fear Europe is being disadvantaged in defending against AI-powered cyber attacks.