German officials warn Anthropic's Mythos AI could facilitate cyberattacks
German government security and financial officials warned in May 2026 that Anthropic's Mythos AI model could facilitate cyberattacks by quickly identifying and exploiting software vulnerabilities, including those in financial institution systems. Officials called for comprehensive security testing and governance measures before deploying such models.
What happened
Stephan Kramer, president of Thuringia's Constitutional Protection Authority, warned that Anthropic's AI model 'Mythos' lowers the barrier for offensive cyber operations because it can quickly find and partially exploit software vulnerabilities.
Kramer cited reports that Mythos had partially autonomously completed a complex 32-step cyberattack simulation in tests.
Joachim Nagel, president of the Bundesbank, warned in April that Mythos can quickly identify and exploit security gaps in financial institution software, presenting a double-edged sword that could be misused for malicious purposes.
Reported impact
Kramer stated the risk is particularly serious if criminals or state actors gain access to the model.
Nagel warned that the technology could be used to exploit weaknesses in financial systems for malicious purposes.
Index analysis
Kramer called for treating such models as 'strategic risk technology' in Europe and recommended comprehensive security testing, licensing procedures, and red-teaming tests before release.
Kramer also demanded clear liability rules for providers and closer collaboration between European regulatory authorities.
Governance controls
Nagel stated that all relevant institutions must have access to the technology to avoid competitive distortions, while emphasizing the need to prevent misuse.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.