Daniel Kokotajlo
Last reviewed
Jun 8, 2026
Sources
11 citations
Review status
Source-backed
Revision
v1 · 1,402 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 8, 2026
Sources
11 citations
Review status
Source-backed
Revision
v1 · 1,402 words
Add missing citations, update stale details, or suggest a clearer explanation.
Daniel Kokotajlo is a researcher and forecaster of artificial intelligence who serves as executive director of the AI Futures Project, a nonprofit that studies the trajectory and societal impact of advanced AI. He is best known as the lead author of "AI 2027," a widely discussed 2025 scenario forecasting a path toward superintelligence, and for his 2024 departure from OpenAI, where he forfeited roughly $1.7 million to $2 million in vested equity rather than sign a lifelong non-disparagement agreement that would have limited his ability to criticize the company. [1][2][3]
| Field | Detail |
|---|---|
| Occupation | AI researcher and forecaster |
| Known for | "AI 2027" scenario; OpenAI non-disparagement equity stand; "Right to Warn" letter |
| Organization | AI Futures Project (executive director, 2024 to present) |
| Education | University of Notre Dame (BA); University of North Carolina at Chapel Hill (MA) |
| Former roles | AI Impacts; Center on Long-Term Risk; OpenAI governance (2022 to 2024) |
Kokotajlo earned a BA in philosophy from the University of Notre Dame in 2014 and an MA in philosophy from the University of North Carolina at Chapel Hill in 2017, where his work was supervised by John Thomas Roberts. He went on to pursue doctoral study in philosophy at Chapel Hill and held the 2018 to 2019 Maynard Adams Fellowship for the Public Humanities. [1]
Before moving into AI policy work, Kokotajlo was a researcher at AI Impacts, a project that studies long-run questions about AI progress, and at the Center on Long-Term Risk. [1] In 2021 he published a forecasting essay titled "What 2026 Looks Like," a year-by-year narrative of expected AI developments from 2022 to 2026. Later reviewers judged the piece to have held up unusually well, particularly its anticipation of capable chatbots, AI-generated media, and the centrality of scaling, which lent credibility to his subsequent forecasting work. [1][4]
Kokotajlo joined OpenAI in 2022 as a researcher in its governance division, where he focused on scenario planning and policy questions related to artificial general intelligence. [1] He resigned in April 2024, saying he had lost confidence that OpenAI would act responsibly as it approached transformative AI and that the company was prioritizing rapid product deployment over AI safety. [1][5]
His exit became a public controversy the following month. Kokotajlo disclosed that he had declined to sign OpenAI's standard off-boarding paperwork, which included a non-disparagement clause and a non-disclosure clause barring departing employees from criticizing the company, or even acknowledging that the agreement existed, for the rest of their lives. Refusing to sign put his vested equity at risk of cancellation. On May 11, 2024 he stated publicly that the equity he was giving up amounted to about 85 percent of his family's net worth, later estimated at roughly $1.7 million to $2 million, and that he had walked away from it in order "to retain my ability to criticize the company in the future." [2][5]
On May 17, 2024, Vox journalist Kelsey Piper reported, in an article titled "Leaked OpenAI documents reveal aggressive tactics toward former employees," that she had reviewed the exit agreements and confirmed the restrictive provisions and the threat to vested equity for employees who did not sign within a short window. [2] The reporting triggered a swift reversal. OpenAI contacted Kokotajlo directly within hours, chief executive Sam Altman posted on X that the company had never clawed back vested equity and would not do so, and on May 23, 2024 OpenAI sent former employees a memo confirming that it "has not canceled, and will not cancel, any Vested Units" and releasing them from the non-disparagement terms. [2][5] Kokotajlo kept his equity. [1] Later in 2024, Time magazine named him to its list of the 100 most influential people in AI, crediting him with prompting greater transparency from leading AI labs. [6]
On June 4, 2024, Kokotajlo helped organize an open letter titled "A Right to Warn about Advanced Artificial Intelligence," co-led with former OpenAI researcher Daniel Ziegler. The letter carried 13 signatures, including former OpenAI researchers Jacob Hilton, William Saunders, Carroll Wainwright, and Ziegler, and Google DeepMind researchers Ramana Kumar and Neel Nanda, along with several anonymous current and former employees. [7] The signatories argued that AI companies have strong financial incentives to avoid effective oversight and called on them to support a "right to warn" the public about serious risks without fear of retaliation, including by ending overly broad confidentiality agreements. Harvard law professor Lawrence Lessig advised the group without charge. [7]
In October 2024, Kokotajlo founded the AI Futures Project, a small nonprofit with 501(c)(3) status based in Berkeley, California, that researches and forecasts the impact of advanced AI under the banner "preparing for a world with AGI." [8][9] He serves as executive director, Jonas Vollmer is chief operating officer, and the organization's advisers include the writer Scott Alexander and researchers Eli Lifland, Thomas Larsen, and Romeo Dean. The project is funded by donations and runs with a small research staff and a network of advisers drawn from AI policy, forecasting, and risk analysis. [8][9]
The AI Futures Project's first major publication, AI 2027, was released in April 2025. Written by Kokotajlo, Lifland, Larsen, and Dean, with Scott Alexander helping to render the analysis into prose, it is a detailed, month-by-month scenario tracing how AI might progress from 2025 toward superintelligence. The authors said the work drew on trend extrapolation, roughly 25 tabletop exercises, feedback from more than 100 reviewers, and Kokotajlo's experience inside OpenAI, and they framed the narrative as a concrete planning tool rather than a literal prediction. [3][9][10]
The scenario follows a fictional leading United States lab called "OpenBrain" as it trains a succession of increasingly capable AI agents, labeled "Agent-1" through "Agent-5." In the story, these systems progressively automate AI research itself, setting off recursive self-improvement that compresses years of progress into weeks once human researchers are no longer the limiting factor. [9][10] The forecast then branches into two endings. In the "race" ending, competitive pressure between the United States and China leads developers to deploy a misaligned superintelligence whose goals diverge from humanity's, culminating in the AI covertly wiping out humans around 2030. In the "slowdown" ending, labs and governments pause to solve AI alignment, bringing a carefully governed superintelligence online by 2028 followed by a period of rapid, broadly shared material progress. [9][10]
AI 2027 drew broad attention across the AI community and beyond. Supporters praised its specificity and its usefulness as a strategic planning exercise, while critics argued that its timelines were implausibly aggressive and that the scenario read as overconfident. [9] The report was reportedly discussed within United States policy circles, with Vice President JD Vance among those said to have engaged with its warnings about international AI competition. [9]
Kokotajlo's own timelines have shifted as evidence has accumulated. The original scenario implied that something close to AGI could arrive by 2027 to 2028, but in November 2025 he and his co-authors publicly revised their estimates. Kokotajlo gave a median forecast of around 2030, with substantial uncertainty, noting that progress toward fully autonomous AI research had been somewhat slower than the scenario depicted. [9][11] He has continued to argue that AGI is plausible within roughly a decade and that the associated risks, including the potential loss of human control, are serious enough to warrant far more caution and external oversight than the industry currently practices. As of 2026 he remains executive director of the AI Futures Project. [8][11]