The German banking sector does not view Anthropic's new AI model, Mythos, as an existential threat, despite concerns about its cyber capabilities. This was stated by Deutsche Bank CEO Christian Sewing in an interview with Bloomberg.
“This is certainly not a reason for panic or concern on our part. However, it is something we must take into account in our daily risk management,” said Sewing, who also chairs the Association of German Banks.
According to him, the organization has formed a working group to provide information and recommendations primarily to smaller financial institutions. Sewing emphasized that European banks have made “huge strides” in enhancing their cybersecurity in recent years.
What Happened
In early April, Anthropic unveiled its new model, Claude Mythos, but refused to release it publicly, citing high security risks. Instead of a public launch, the company initiated Project Glasswing—a collaboration involving Amazon, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, Linux Foundation, Microsoft, Nvidia, and Palo Alto Networks to test the tool in secure environments.
“AI models have reached a level of programming skills that allows them to surpass all but the most qualified humans in identifying and exploiting software vulnerabilities,” stated Anthropic.
The skills demonstrated by Mythos in tests have raised concerns among regulators worldwide. In the UK and Canada, central banks held meetings with representatives from relevant agencies and major businesses. U.S. Treasury Secretary Scott Bessen and Federal Reserve Chair Jerome Powell convened an emergency meeting with Wall Street executives.
Financial firms in other regions are also urging Anthropic to allow them to test Mythos within their own systems.
“Naturally, everyone is trying to gain access, but I believe it is entirely appropriate that it remains limited for now. This ensures we do not inadvertently slide into a situation of overgeneralization that could potentially exacerbate the problem,” Sewing remarked.
Anthropic and the U.S. Administration
On February 27, U.S. President Donald Trump ordered all federal agencies to completely cease using Anthropic technologies within six months.
“We don’t need it, we don’t want it, and we will no longer do business with them!” the president stated.
Earlier in January, the WSJ reported on the risk of the Department of Defense terminating its contract with Anthropic, awarded in July 2025. Disagreements arose due to the startup's strict ethical policies, which prohibit using the Claude model for mass surveillance and autonomous lethal operations.
Officials' dissatisfaction grew amid the integration of the Grok chatbot into the Pentagon's network. Anthropic's attempt to adjust its AI safety rules to meet military requirements did not help either.
However, following the emergence of the Mythos situation in April, the Trump administration and Anthropic CEO Dario Amodei discussed the possibility of continued collaboration. According to Reuters, the meeting included White House Chief of Staff Suzy Wiles and Treasury Secretary Bessen.
“We discussed opportunities for collaboration, as well as common approaches and protocols to address issues related to scaling this technology,” the administration stated.
In April, Anthropic introduced Claude Opus 4.7 for advanced development and launched an experimental design tool called Claude Design.