The European Commission is currently in talks with OpenAI and Anthropic regarding access to their artificial intelligence models. A spokesperson indicated that OpenAI has proactively offered to provide access to its new AI model, which is viewed positively by the Commission; in contrast, discussions with Anthropic have not yet progressed to potential access points, despite four or five meetings that have focused on good exchanges. These negotiations occur in the broader context of the EU’s recent efforts to streamline AI regulations, particularly concerning high-risk systems, while also emphasizing cybersecurity features in AI models, including those being developed by both companies.

OpenAI: OpenAI is a U.S.-based artificial intelligence company developing advanced large language models and specialized tools for various applications including cybersecurity. It recently updated its principles to address societal adaptations to AI advancements. In the news, OpenAI is proactively engaging with the EU Commission, offering access to its new cybersecurity AI model during these talks.
Anthropic: Anthropic is a U.S. AI company focused on safety-aligned models like Claude and newer developments such as the Mythos model with cybersecurity capabilities. It is expanding data center presence in Europe to support its AI infrastructure needs. Regarding the EU talks, Anthropic has participated in several meetings with the Commission about its AI models but has not yet progressed to offering access.
EU Commission: The European Commission acts as the EU’s executive arm, proposing legislation and overseeing enforcement of regulations like the AI Act to manage emerging technologies. Recently, it agreed with lawmakers to simplify AI rules and ban certain high-risk applications such as nudifier apps to foster innovation while protecting citizens. In this context, it is holding ongoing discussions with OpenAI and Anthropic to gain insights into and potential access to their advanced AI models, particularly cybersecurity-focused ones.

`json
{
“Negotiation Stages”: “OpenAI has offered model access to the EU, while Anthropic discussions remain at an earlier exploratory phase.”,
“Cybersecurity Emphasis”: “Talks involve AI models, but the specific features or models discussed are not detailed in the available news.”
}
`