Several high-profile figures and teams in the AI industry have publicly criticized their former employers’ approach to safety, oversight, and how risks are handled as models advance. The latest flashpoint involves OpenAI firing three safety and alignment researchers—Mikita Balesni, Tomek Korbak, and Jasmine Wang—after they said they were removed for speaking out about AI safety. OpenAI counters that an internal investigation found they violated policies on handling sensitive information.

Across reporting, the broader pattern extends beyond OpenAI. Fast Company describes a growing list of insiders leaving AI labs after raising concerns about issues such as copyright practices, inadequate safety culture, and the pace of model development. Examples include David Robinson’s resignation and call to end “trial and error,” Jan Leike’s departure and warning that safety is taking a back seat to product launches, and a June letter signed by dozens of current and former AI staff and prominent researchers calling for a “right to warn” about advanced AI risks.

Outlets also note additional dissent around military use of AI and concerns over misinformation and third-party oversight. The accounts converge on one theme: tensions persist between institutional processes for safety work and the way some employees believe risks are communicated—or dismissed—inside major AI companies.