Staff at leading AI companies are experiencing significant mental strain from fears their work could harm society. Multiple employees at the UK's AI Safety Institute have been signed off work with stress, citing tight testing schedules and alarm over rapid AI deployment. Similar issues have been reported at Anthropic, Open AI, and Google Deep Mind, with many top researchers quitting due to mental toll or safety concerns. A Deep Mind employee described the work as 'an impossible task' and warned of burnout risks from uncertainty about how AI systems will be used, particularly after Google's deal with the Pentagon for military AI applications.

Former AI Safety Institute researchers highlighted stress from working in cyber security and biochemistry, where AI capabilities now allow complex tasks like finding code flaws and creating unknown viruses. One warned of 'constant anxiety' about whether anything can stop reckless AI development. Another described a 'very stressful and at times toxic management and org culture' despite believing the work tackles the world's most important problem.

The UK government and Deep Mind both emphasized wellbeing support, with Deep Mind stating researcher safety is their 'top priority' and offering comprehensive well-being resources and clear channels for escalating concerns.