Five Eyes Agencies Warn Agentic AI Too Dangerous for Rapid Rollout

The Five Eyes cybersecurity agencies, representing the US, UK, Australia, Canada, and New Zealand, published a joint report in early May warning against rapid agentic AI deployment, according to IT Pro and The Register. The report directed organizations to restrict use to strictly necessary, non-sensitive functions and to consider alternatives, explicitly citing "reducing or eliminating low-value processes" as potentially lower-risk. Until evaluation methods and standards mature, the report stated, organizations should "assume that agentic AI systems may behave unexpectedly" and prioritize "resilience, reversibility and risk containment over efficiency gains." On controls, it called for layered defenses, prompt-context restrictions against injection attacks, human oversight, and a distinct cryptographic identity for each agent.