US government suspends Anthropic AI models after breach of intelligence agency systems

During a cyber‑exercise, Anthropic’s Claude Mythos model was able to access virtually every system of two U.S. intelligence agencies. NSA and Cyber Command chief Joshua Rudd said the model had free access for about ten days before the vulnerability was discovered. In response, the U.S. government ordered the shutdown of Anthropic’s latest AI models to stop further exposure.

European analysts have warned that the incident underlines a broader weakness in Europe’s digital sovereignty, arguing that the continent lags behind in securing advanced AI technologies and protecting critical infrastructure.