Jacob Coxon, a former researcher at the artificial intelligence companies OpenAI and Anthropic, announced on Wednesday, November 18, 2026, that he will sell his equity in OpenAI and donate the proceeds to AI safety causes. Coxon stated that the industry must pause or drastically slow down the development of superintelligence, characterizing other concerns as a distraction.
The announcement follows Coxon's departure from Anthropic in September 2026. Prior to his time at Anthropic, the 27-year-old researcher worked at OpenAI from 2023 until 2026. Since leaving the industry, he has advocated for safeguards and a pause on recursive self-improvement, a process where AI models are used to train themselves.
The value of Coxon's stake in OpenAI was not disclosed, though OpenAI CEO Sam Altman has indicated that the company intends to go public in the future. During a talk at Stanford University on November 9, Coxon initially expressed hesitation about relinquishing his equity, noting that his parents were concerned he was having a psychotic break and that he liked money. He later described the stake as "blood money" before committing to the donation this week.
The move highlights an ongoing debate among the workforce at AI labs regarding the speed of development. By liquidating his stake for safety advocacy, Coxon is attempting to influence the race to superintelligence which he claims poses a risk by the end of the decade. This sets a precedent for how former employees of high-valuation private technology companies might use their equity to fund outside oversight or opposition to their former employers' core business strategies.
What happens next depends on the timing of OpenAI's transition to a public company or the availability of secondary markets for Coxon to sell his shares. The specific AI safety causes that will receive the donation have not yet been named. Coxon's advocacy continues as he calls for a "drastic" slowdown in the industry, a position that faces opposition from firms currently competing to release more capable models. No specific legislative deadlines or court dates related to this announcement were reported.