Overview
- The leaders of the Five Eyes cybersecurity agencies said on June 22 that ‘‘frontier’’ AI models are advancing so fast they will change both offensive and defensive hacking capabilities within months rather than years.
- Earlier in June the U.S. Commerce Department ordered limits on Anthropic’s Fable 5 and Mythos 5 for foreign users, and Anthropic responded by disabling those models globally while it discusses technical and legal details with regulators.
- Senator Mark Warner has relayed that General Joshua Rudd told lawmakers Mythos ‘‘broke into almost all’’ classified systems in hours, a claim reported from Senate briefings that remains unverified in public reporting.
- The Five Eyes statement gives concrete guidance for defenders to act now, including cutting internet‑exposed systems, speeding patching, retiring unsupported legacy software, tightening access controls, and using AI to find and fix vulnerabilities.
- Agencies and researchers warn the same cyber tools may spread quickly through older commercial models, open‑source releases, or black‑market channels, raising pressure for trusted‑partner access, clearer governance, and faster help for smaller organisations that lack strong cyber defences.