Anthropic Provides Access to Mythos AI Model for the EU
US AI company Anthropic is now providing access to its powerful AI model, Mythos, to the European Union.
This move is part of the Project Glasswing expansion plan, with ENISA being the first EU agency to gain access for defensive testing.
Market Mechanism: EU agencies, as buyers, receive restricted access, shifting capital and technology collaboration from US dominance to transatlantic sharing. Anthropic benefits from expanding its strategic partner network and alleviating regulatory pressure, while European AI and cybersecurity firms enhance their capabilities, reducing reliance on US technology.
Supplementary Data: Mythos excels at discovering and exploiting zero-day vulnerabilities, achieving an 83.1% success rate in the CyberGym benchmark test.
Source: Public Information
ABAB AI Insight
Anthropic previously restricted access to Mythos (Claude Mythos Preview) exclusively to US companies and government entities. This extension to ENISA continues its Project Glasswing initiative, which has managed the high-risk capabilities of the Claude series through similar controlled releases in the past.
On the capital front, Anthropic chose limited openness after pressure and negotiations from the EU, motivated by the desire to reduce geopolitical regulatory risks through defensive authorizations while maintaining core control of the model. This paves the way for future collaboration and revenue with European firms, responding to competition from OpenAI, which has also opened similar models to the EU.
Similar to OpenAI's recent provision of cybersecurity model access to the EU, Anthropic is currently transitioning from a US-centric deployment to limited expansion among allies, focusing on balancing security control with global influence.
Structural Judgment: This fundamentally reflects regulatory changes. Under continued pressure from the EU, Anthropic is adjusting its access policy, achieving transatlantic technology sharing by opening access to ENISA, and promoting a shift in pricing power from unilateral US control to multilateral regulatory coordination. The mechanism is driven by the global diffusion pressure of high-risk AI models, forcing private companies to accept intergovernmental limited access frameworks.
ABAB News · Cognitive Law
Access to powerful AI has transformed from a national security issue into a diplomatic bargaining chip. The longer the refusal to open access, the more external pressure translates into long-term dependency. The real scarcity is not the model itself, but who can legally use it first.