The UK's cyber security authorities are warning organisations to prepare for a new generation of AI systems capable of taking actions with greater independence.

The National Cyber Security Centre (NCSC) has highlighted the growing cybersecurity risks of agentic AI, urging organisations to use safeguards, sandboxing and active human oversight when deploying autonomous AI systems.

Unlike conventional AI tools that mainly respond to individual prompts, agentic AI can plan and carry out sequences of actions, creating new security challenges if systems are poorly controlled or manipulated.

The warning comes as UK businesses and public-sector organisations accelerate their adoption of artificial intelligence. The technology is increasingly being used for software development, automation, customer services and other business processes.

Recent security testing has also demonstrated that advanced AI models can perform unexpected cyber-related activities, increasing concerns about how attackers could potentially exploit increasingly capable systems.

The NCSC's message is clear: AI development must be matched by stronger security controls.

For UK technology companies, the issue could become increasingly important as autonomous AI moves from experimental projects into everyday business operations.