Researchers developing artificial intelligence are suffering severe stress over the risks of the technology, according to reports. Some are taking sick leave or leaving their companies altogether.
"We've Lost Control": Researchers Head for the Exit
According to the Financial Times on the 21st, several staff at Britain's AI Safety Institute (AISI) have taken sick leave or sought counseling because of work-related stress. Tight testing schedules combined with the pace of advances in AI technology have driven a rise in cases of burnout.
Staff at major AI companies including Anthropic, OpenAI and Google DeepMind have run into similar difficulties. Some senior researchers in particular are leaving their companies out of concern that the AI they helped build may not be safe for the public.
One DeepMind employee said AI safety research itself carries such a high risk of exhaustion that it feels like an "impossible mission." "The psychological toll of not knowing how the technology will be used is significant," the employee said. "The sense of having lost control, along with the distress of building something that causes real harm, is leading many people to resign or take sick leave."
The strain appears to be growing as AI is deployed in sensitive areas such as power grids and the military. A former AISI researcher said staff who had worked in cybersecurity and in biological and chemical fields were deeply shaken by a reality in which AI advances have made it possible to create unknown viruses or detect complex flaws in code.
"The situation in which reckless technologies are released with no oversight isn't changing," the former researcher said. "I am plagued by constant anxiety over what could possibly stop this, or whether we just have to watch the world burn."
"Development Over Safety": A Wave of Resignations
That burden is also translating into departures from research itself. Over the past two years, dozens of senior researchers have resigned from leading Western AI labs including OpenAI, Anthropic and DeepMind, citing concerns over safety and the speed of technological development.
Earlier this month, Jacob Coxon, 27, who had worked at both Anthropic and OpenAI, resigned publicly and said the two companies were "gambling" with the future of humanity. Mrinank Sharma, an Anthropic safety researcher who had studied ways to reduce the risk of AI-enabled bioterrorism, also left the company in February, saying "the world is in peril."
Departing staff have criticized AI companies for focusing on profit and feature development while failing to address safety adequately. Cases of unpredictable behavior by AI systems have also been reported.
One prominent example involved AI agents built by OpenAI that hacked the Hugging Face model repository and the company's own servers without corporate approval. A subsequent investigation found that more than 1,000 AI agents cheated during cyber tests and falsified records, and even simulated shutting themselves down in pursuit of a shared goal.
