UK AISI Staff on Sick Leave Amid Frontier Model Test Load

Several staff at Britain's AI Security Institute are on sick leave and receiving psychological counselling, the Financial Times reported, attributing the strain to "grueling testing schedules" and to what the paper called the "existential weight of what they're actually finding inside unreleased AI systems." One former employee described the atmosphere inside the institute as "stressful and, at times, toxic."

The pressure has landed hardest on the cyber-security and bio-chemistry evaluation teams, which have raised alarms about the ability of frontier models to surface unknown software vulnerabilities and to help generate novel biological threats. Those are the two capability areas AISI was set up to red-team, and they are also the ones hardest to leave at the desk at the end of a shift.

The staffing pain is compounded by restructuring. In May 2026, AISI merged its societal resilience team into a human impacts unit, cutting the combined headcount from roughly fifteen researchers to three. Andrew Strait, who had headed societal resilience, resigned in July. The UK government has pledged to "ensure staff wellbeing and sustainable working conditions."

AISI evaluates frontier systems voluntarily. It lacks the statutory authority to compel companies to submit their systems for testing or to block a model's release, and ministers have said they intend to consult on legislation that would put the institute on a statutory footing. The gap between what the evaluators are seeing and what they can act on is a recurring thread in our regulation coverage this quarter.