Anthropic joins Mythos in knocking on the EU's door: Commitment to follow AI guidelines, internet security model triggers strict regulations.

Anthropic is in discussions with the European Commission regarding the deployment of multiple AI models in the EU, including a cybersecurity model that has not yet been launched locally. The company has committed to adhering to the EU's general AI practices guidelines and has obligations to assess and mitigate potential risks. European Commission spokesperson Thomas Reynier stated that Anthropic has committed to adhering to the EU's general AI practices guidelines. Reynier pointed out, "Within this framework, companies have an obligation to assess and mitigate risks that may arise from a service (whether or not it is offered in Europe)." Neither Anthropic nor Reynier immediately responded to requests for comments. Last week, Anthropic released its most powerful model to date, Claude Mythos Preview, which focuses on programming and intelligent agent task abilities. Despite legal disputes with the Trump administration due to being blacklisted, the company is still in communication with the US regarding Mythos. CEO Dario Amodei is scheduled to meet with the White House Chief of Staff, potentially bringing a breakthrough to resolve the AI usage dispute with the Pentagon. Mythos sparks cybersecurity concerns across nations Following the release of Mythos, US Treasury Secretary Benson and Federal Reserve Chairman Powell urgently convened a meeting on April 7 with CEOs of major banks including Citigroup, Morgan Stanley, Bank of America, Vanguard, and Goldman Sachs to discuss the "new era of network risk" that Mythos could potentially open. All banks summoned are identified as "systemically important financial institutions," and their stability is crucial for the global financial system. The UK is also highly anxious. Bank of England Governor Bailey publicly stated, "One would think that events in the Gulf region are the latest challenges facing the world. But waking up to find that Anthropic may have found a way to completely disrupt the cybersecurity landscape, you can't help but wonder what did I do wrong in my past life." Why financial regulators are so nervous? The key lies in the threat Mythos poses to the SEC's "comprehensive audit tracking" database. The American Securities Association (ASA) pointed out in a letter to Benson that Mythos could be exploited by criminals to attack the database, leading to large-scale identity theft, exposure of personal transaction holdings, and amplification of internal threats. More profoundly, the core systems of large global banks mostly rely on homogenized underlying software, and once Mythos finds a vulnerability in a common component, the attack could spread like dominoes, directly threatening cross-border payments, clearing settlements, and other financial core links. Window for reconciliation with the US government opens? Previously, the US Department of Defense terminated its contract with Anthropic and labeled it as a supply chain risk company, leading Anthropic to file a lawsuit against the Department of Defense. The manufacturer of Claude AI chatbot Siasun Robot & Automation, supported by Amazon and Alphabet's Google, refused the Department of Defense's demand for unrestricted access to its AI models. On April 17, 2026, Amodei will meet with White House Chief of Staff Susie Wells, which is seen by the public as an important signal that a breakthrough in the dispute between the two sides may be achieved. The Trump administration has recognized the advanced capabilities of Anthropic's new AI model "Mythos" in cybersecurity defense and is considering granting federal agencies access to it. Anthropic also emphasizes that its position is not to completely reject cooperation with the military, but to advance within a reasonable framework. In a statement following the latest lawsuit ruling, the company stated, "While lawsuits are necessary to protect the company, customers, and partners, we remain committed to engaging in constructive cooperation with the government to ensure all Americans benefit from safe and reliable artificial intelligence."