EU cybersecurity agency granted access to Mythos 5 AI model, Commission says

• ENISA is currently testing Anthropic's Mythos 5 AI model after receiving access approval
• The cybersecurity agency also gained access to OpenAI's latest GPT-6-Astra model
• This represents the EU's effort to evaluate and assess advanced AI systems for security and regulatory purposes

EU Cybersecurity Agency Gains Access to Major AI Models

The European Union's cybersecurity agency ENISA has been granted access to two significant artificial intelligence models for testing purposes, according to a European Commission spokesman on September 10, 2026.

Key Developments:

ENISA now has access to:

• Anthropic's Mythos 5 AI model
• OpenAI's latest model, GPT-6-Astra

The agency is currently conducting testing on these systems, though specific details about the scope and objectives of the testing were not disclosed.

Market Implications:

This development signals increased regulatory scrutiny of advanced AI systems within the European Union. The move appears consistent with the EU's proactive approach to technology regulation and cybersecurity oversight, particularly regarding cutting-edge AI models from major U.S.-based companies.

The access granted to ENISA suggests potential compliance requirements or security assessments that AI developers may need to satisfy to operate within EU markets. This could have implications for:

• AI model deployment timelines in Europe
• Regulatory compliance costs for AI companies
• Competitive dynamics between AI providers operating in the EU market
• Future EU policy framework for advanced AI systems

Companies Involved:

• Anthropic: Provider of Mythos 5 model
• OpenAI: Provider of GPT-6-Astra model

The announcement reflects the EU's continued focus on balancing technological innovation with security and regulatory oversight, potentially setting precedents for how advanced AI models are vetted before broader deployment in European markets. No specific timelines for the completion of testing or subsequent actions were provided.