Timeline
- 7.0
Fired OpenAI safety researchers dispute misconduct claims, warn of chilling effect
Three fired OpenAI safety researchers publicly rejected allegations that they mishandled sensitive information, publishing an open letter that disputes the misconduct claims and warns their dismissals are chilling AI safety work inside the company. In the letter they wrote that "given the significant safety concerns surrounding the development of AI, employees must not be left working in an environment where fear and unclear rules stymie AI safety work and weaken third-party accountability." The letter frames the dismissals as a governance issue rather than a purely personnel matter, arguing that the way the cases were handled affects how safety work and third-party accountability function at a leading AI lab. The researchers dispute the characterization that they improperly shared sensitive information; earlier reporting linked the dismissals to allegations of sharing information with external safety groups.
- 7.0
An OpenAI safety employee has quit and is sounding the alarm
David Robinson, who wrote the safety reports accompanying OpenAI's major model releases, resigned this week and is publicly warning about AI safety risks in an editorial published by The Atlantic. According to reports on his departure, he spent about three and a half years at OpenAI, led the drafting of its current Preparedness Framework, and oversaw safety reporting on 12 frontier launches. Robinson authored the safety documentation that accompanied OpenAI's major model releases, so his public criticism adds to outside scrutiny of how frontier labs conduct and disclose safety oversight. Reports on his exit say he wants AI labs to draw lessons from aviation and nuclear safety practices. Robinson ran OpenAI's safety transparency work and took part in developing the system cards that describe its models, according to reports. One report headline described him as calling OpenAI's culture "broken," a characterization attributed to him in press coverage.