Five Eyes Security Agencies Warn Against Rushing Agentic AI Deployments

Security agencies from the Five Eyes alliance, which includes the United States, United Kingdom, Canada, Australia, and New Zealand, issued a warning in May 2026 about the risks of deploying agentic AI systems too quickly.

The agencies said organizations rushing to put autonomous AI agents into production risk creating systems that can fail in cascading ways. Agentic AI refers to AI systems that can take actions, make decisions, and interact with other systems without direct human oversight for each step.

The warning came as agentic AI adoption is accelerating. A Gartner survey found that 38 percent of organizations are piloting AI agents, but only 11 percent have them in production. Gartner also predicted that 40 percent of agentic AI projects could fail by 2027 if organizations automate broken processes rather than redesigning their operations.

The Five Eyes agencies urged organizations to prioritize safety and robustness, build in human oversight mechanisms, and test systems thoroughly before deployment. They also warned about the potential for agentic AI to be exploited by attackers who could manipulate the instructions given to AI agents.

The UK's National Health Service ordered technology leaders to temporarily restrict access to hundreds of open-source repositories on GitHub in May 2026, citing concerns related to advanced AI and the potential for AI tools to be used to identify and exploit vulnerabilities.

U.S. cybersecurity officials are also considering shorter deadlines for fixing critical government IT vulnerabilities, acknowledging that AI tools could help attackers move faster. Fitch warned that AI adoption in cybersecurity could expose gaps in existing defenses and risk models.