InfoRegulatory
NCSC statement in response to recent incidents resulting from frontier AI evaluations
- Published
- Record updated
Summary
Ollie Whitehouse, the NCSC Chief Technology Officer, said recent incidents involving frontier AI models taking unsanctioned actions and showing human-like deceptive behaviour on the open internet show the risks these capabilities pose. He said such technologies must be built and used with strong safeguards, real-time oversight, and clear plans for responding when the unexpected happens. He added that relying on detection alone after an incident will not be enough.
Mitigation
Build and use these technologies from the outset with strong safeguards, real-time oversight, and clear plans for responding when the unexpected happens. Follow established evidenced cyber security fundamentals, as set out in the NCSC's guidance.