Home
Anthropic Grants EU Access to Top-Security Model Mythos After Months of Talks, Defying US Opposition

There has been notable progress in global governance of AI security and geopolitical diplomacy. On Monday, the European Commission officially confirmed that after months of diplomatic efforts and several productive meetings, the U.S. leading AI startup Anthropic has decided to grant the EU access to its most advanced AI security model, "Mythos." This initiative aims to help the EU better evaluate and understand the potential cybersecurity risks posed by next-generation cutting-edge large models.
As a core component of Anthropic's cybersecurity project "Project Glasswing," the Mythos model was initially made available to a select group of companies in April this year. The model boasts strong specialized security analysis capabilities, accurately identifying thousands of previously unknown security vulnerabilities and weaknesses in complex software. However, this powerful "double-edged sword" attribute raised deep concerns among governments, financial institutions, and tech giants upon its release. There is widespread fear that if this technology falls into the hands of malicious actors, it could significantly accelerate cybercrime and create serious cybersecurity threats.
Behind this battle over access rights, the process has been full of twists and turns. Previously, insiders revealed that due to the extremely high technical secrets and strategic value involved, Anthropic had informed the EU that it needed permission from the U.S. government. In response, the EU has intensified communication with the U.S. administration over the past week. The U.S. government generally opposes sharing such core underlying models with non-U.S. official organizations, aiming to maintain the United States' absolute dominance and technological sovereignty in the global AI field.
Despite these geopolitical technological barriers, the EU continues to accelerate breakthroughs in promoting advanced AI regulation and security collaboration. Prior to this, the EU successfully obtained access to the advanced security model GPT-5.5-Cyber from OpenAI in May this year. At the same time, safety pre-review of large models is becoming a mainstream global trend. In May this year, U.S. relevant government departments also reached similar agreements with Google DeepMind, Microsoft, and xAI, Musk's company, allowing official entities to conduct comprehensive security and risk assessments of AI models before their public release.
Thomas Regnier, the EU spokesperson for technological sovereignty, welcomed this development in a statement, emphasizing that as a new wave of more powerful models accelerates into the market, balancing technological innovation and cybersecurity has become a serious challenge for the international community. Currently, the EU is further strengthening dialogue with like-minded partners, including the United States, in an attempt to build more normalized security defenses in the core areas of the global AI value chain.
Related article
DeepMind CEO Hassabis: I sleep six hours a day, usually feel energetic around 1 a.m.
Fortune recently featured an interview with Demis Hassabis, CEO of Google DeepMind, revealing his unconventional approach to rest and productivity. Hassabis disclosed that he sleeps very little, structuring his waking hours into two distinct work blo
OpenAI, Anthropic Vie for Market Share Despite Revenue Shortfalls
Despite recent reports suggesting OpenAI missed revenue targets, creating pressure on tech stocks this Tuesday, private AI lab investors remain resilient. Seasoned backers have confirmed they will not reduce investment despite negative media coverage
California AV Compliance: A New Era of Tickets, Geofences, and 1M Miles
Guident operates an AuveTech shuttle in South Florida, managing a four-mile route in West Palm Beach and a one-mile route in Boca Raton using its remote monitoring technology. | Credit: GuidentCalifornia is redefining the regulatory landscape for dri
Related Special Topic Recommendations
Comments (0)
0/500

There has been notable progress in global governance of AI security and geopolitical diplomacy. On Monday, the European Commission officially confirmed that after months of diplomatic efforts and several productive meetings, the U.S. leading AI startup
As a core component of Anthropic's cybersecurity project "Project Glasswing," the Mythos model was initially made available to a select group of companies in April this year. The model boasts strong specialized security analysis capabilities, accurately identifying thousands of previously unknown security vulnerabilities and weaknesses in complex software. However, this powerful "double-edged sword" attribute raised deep concerns among governments, financial institutions, and tech giants upon its release. There is widespread fear that if this technology falls into the hands of malicious actors, it could significantly accelerate cybercrime and create serious cybersecurity threats.
Behind this battle over access rights, the process has been full of twists and turns. Previously, insiders revealed that due to the extremely high technical secrets and strategic value involved, Anthropic had informed the EU that it needed permission from the U.S. government. In response, the EU has intensified communication with the U.S. administration over the past week. The U.S. government generally opposes sharing such core underlying models with non-U.S. official organizations, aiming to maintain the United States' absolute dominance and technological sovereignty in the global AI field.
Despite these geopolitical technological barriers, the EU continues to accelerate breakthroughs in promoting advanced AI regulation and security collaboration. Prior to this, the EU successfully obtained access to the advanced security model GPT-5.5-Cyber from
Thomas Regnier, the EU spokesperson for technological sovereignty, welcomed this development in a statement, emphasizing that as a new wave of more powerful models accelerates into the market, balancing technological innovation and cybersecurity has become a serious challenge for the international community. Currently, the EU is further strengthening dialogue with like-minded partners, including the United States, in an attempt to build more normalized security defenses in the core areas of the global AI value chain.
DeepMind CEO Hassabis: I sleep six hours a day, usually feel energetic around 1 a.m.
Fortune recently featured an interview with Demis Hassabis, CEO of Google DeepMind, revealing his unconventional approach to rest and productivity. Hassabis disclosed that he sleeps very little, structuring his waking hours into two distinct work blo
OpenAI, Anthropic Vie for Market Share Despite Revenue Shortfalls
Despite recent reports suggesting OpenAI missed revenue targets, creating pressure on tech stocks this Tuesday, private AI lab investors remain resilient. Seasoned backers have confirmed they will not reduce investment despite negative media coverage











