UK AI Safety Institute staff take stress leave, Financial Times says
Four people told the newspaper that testing schedules for unreleased models and alarm at their capabilities have left morale low and several staff on leave.
Multiple staff at the UK's AI Safety Institute are off work with stress and undergoing counselling, the Financial Times reported on Tuesday. The paper attributed the account to people with knowledge of the matter, and gave no number.
Four people told the FT that tight schedules for testing unreleased models, and alarm at how quickly those models are gaining capability, had contributed to low morale and burnout. The account rests on the FT's reporting alone.
A former researcher at the institute said the cyber and bio teams were the most alarmed by recent jumps in capability. The models can now find undiscovered flaws in code and produce viruses “unknown in nature”, the former researcher said.
A DeepMind employee described “various quittings and a lot of mental health leave”, and said the pace of the work is not compatible with safety. The FT did not name the employee and gave no figure for departures.
The FT describes the institute as the global leader in independent testing of frontier models. That characterisation is the paper's, not a claim from the institute. What the reported strain means for the testing schedule is not addressed in the account.