The German financial regulator, BaFin, has warned about potential risks to banks posed by Anthropic’s new AI model, “Mythos” and similar offerings. Speaking to the “Handelsblatt” (Wednesday edition), a BaFin spokesperson noted that “there have already been other comparable approaches, and it can be expected that other providers will also introduce such models”. The regulator advised that “financial institutions must prepare for the possibility that vulnerabilities could be discovered soon, which then need to be addressed promptly and rapidly”.
Anthropic unveiled its advanced model, Claude Mythos Preview, last week, which has the capability to autonomously and extensively find security gaps within IT systems. Consequently, US Finance Secretary Scott Bessent and Fed Chair Jerome Powell abruptly convened top representatives from Wall Street for a meeting focused on Mythos. The goal of this meeting was to raise awareness among risk institutes regarding the potential of the new AI model and similar systems, ensuring that appropriate protective measures are implemented.
According to financial circles, such an ad-hoc high-level meeting between banks and authorities, as occurred in the United States, has not taken place in Germany or at the European level. Nevertheless, new cybersecurity threats are discussed regularly through exchanges between authorities and banks across Europe. Some major US banks, such as Goldman Sachs, are already utilizing “Mythos” according to the “Handelsblatt”.
In contrast, German banks reportedly do not yet have access to the model. Commerzbank and DZ Bank both informed the newspaper that they are currently not using Mythos. Deutsche Bank declined to comment on “individual providers or specific AI models”.