ChatGPT Whistleblower: They Offered Me $2m To Keep Quiet, No One Is Ready For What’s Coming!

Summary
In this interview on The Diary Of A CEO, host Steven Bartlett speaks with Daniel Kokotajlo, a former OpenAI governance researcher turned whistleblower who currently leads the AI Futures Project. Kokotajlo details his forecasts for the arrival of artificial superintelligence, the internal dynamics and race pressures among frontier AI labs like OpenAI and Anthropic, his forfeit of $2 million in vested equity by refusing an anti-disparagement agreement, and policy proposals outlined in his “AI 2027” and “AI 2040: Plan A” scenarios.

What is shown

• [01:27] Bartlett displays Kokotajlo’s 2021 forecasting report and a large printed board of “AI 2040: PLAN A”.
• [22:18] Physical copies and on-screen graphics of the “AI 2027” scenario forecast document detailing timeline predictions from 2024 to 2027.
• [23:23] Infographic of the “AI 2027 Scenario Forecast” showing projected timelines for automated coding, recursive self-improvement, and compute growth.
• [41:11] Flowchart decision tree paper (“2029: Choose a Path”) showing Plan A (Verified Slowdown), Plan B (Shut it all down), Plan C (Burn the Lead), and Plan D (Race to ASI).
• [62:33] Graphical breakdown of the “Share of Labor Output” in the US under Plan A (human cognitive, human physical, AI, and robots).
• [65:23] Detailed examination of the “2029: Choose a Path” decision chart and the full timeline board for “AI 2040: Plan A”.
• [93:10] Interactive demonstration on the physical timeline mat using miniature props (flags, model robotaxi, drone, rocket, chess piece, and toy check for the “Citizen’s Dividend”).
• [101:45] Visual illustrations depicting conceptual megastructure data centers on the ocean and future human living environments.

Claims & numbers

• Kokotajlo states that his median forecast assigns a 50% chance of superintelligence arriving by roughly 2029 (or slipping to 2028–2030) (the presenter states at [03:25]).
• Kokotajlo claims there is an estimated 70% chance that unguided development leads to catastrophic failure or human extinction (the presenter states at [00:07], [28:46]).
• Kokotajlo notes Anthropic grew from approximately $1 billion annualized run rate to tens of billions within a year (the presenter states at [03:52]; on-screen community note clarifies it reached $47B annualized run rate by mid-2026).
• Kokotajlo states that leaving OpenAI in 2024 without signing an anti-disparagement clause originally meant forfeiting about $2 million in vested equity, representing roughly 80% of his family’s net worth, before public backlash caused OpenAI to retract the clawback policy (the presenter states at [18:24], [19:00]).
• Kokotajlo states that frontier AI models have expanded parameter counts from ~175 billion in 2020 to ~10 trillion by 2026, representing nearly a 100x increase (two orders of magnitude) in model scale (the presenter states at [34:48]–[35:02]).
• Under the proposed “AI 2040: Plan A” timeline:
  • 2029: International agreement between the US, China, and allies to halt training of new frontier models.
  • 2030: Transition to total research transparency in training data centers.
  • 2031: 1/5 of all cognitive labor automated by AI.
  • 2033: “Citizen’s Dividend” distributing cash starting at $25,000/year per American, scaling up toward $10M/year per person by 2040 factoring in economic growth (the presenter states at [86:20]).
  • 2035: Planned pause at top human-expert AI capability before advancing to superintelligence in 2040.

Notable quotes

• “With a 70% chance that this goes horribly wrong, like human extinction.” — Daniel Kokotajlo [00:07]
• “I think you should judge people by their actions, not by their words.” — Daniel Kokotajlo [13:21]
• “If they can automate the code, then they can do their own work better and faster, and accelerate progress.” — Daniel Kokotajlo [19:58]

Assessment
This video is an in-depth long-form interview and podcast discussion rather than a technical product demonstration. While real documents, reports, and scenario boards authored by Kokotajlo are examined and referenced, the broader discussions around AGI and timeline milestones remain speculative forecasts and policy arguments rather than demonstrated software capabilities.

Described by gemini-3.8-flash on 2026-10-15 from the video’s audio and frames.