Daniel Kokotajlo

AI Futures Project · Ex-OpenAI staffer · Former OpenAI researcher · Ex-OpenAI researcher

Titles as the sources gave them. In 13 entries, with 6 checked quotations.

Who they are

Roles and parties from Wikidata and from the sources in the Base, as precise as each gives them. When we compare what groups say, each statement counts in the group the person belonged to that day.

  • Researcher, OpenAIformer · source

Wikidata: Q75738671

What they said

Newest first. Each quotation was found word for word in the source.

“IF (big IF) Anthropic really is imminently about to meet the RSP threshold of 2x acceleration over baseline, then it seems timelines to full RSI and superintelligence may be very short indeed.”
AI #193: Progress May Dramatically AccelerateDon't Worry About the Vase, November 5, 2026. Speaking as: Former AI company staff. As quoted by the source, not their own text. Daniel Kokotajlo, quoted by Zvi Mowshowitz, reacting to the S-1 language.
“he thinks that about a year from now uh they're going to be putting Claude in charge of everything at the company, and there's about a 5% chance that at this point it'll already be plotting against them.”
Daniel Kokotajlo: "Most of the World Just Doesn't Yet Believe That It's Really Going to Happen"ControlAI, October 14, 2026. Speaking as: Former AI company staff. Describing a conversation three months earlier with an unnamed Anthropic alignment researcher.
“the core problem is that most of the world just doesn't yet believe that it's really going to happen.”
Daniel Kokotajlo: "Most of the World Just Doesn't Yet Believe That It's Really Going to Happen"ControlAI, October 14, 2026. Speaking as: Former AI company staff. On what stands in the way of action on AI risk.
“I think that once people realize that that these companies might actually build super intelligence in the next few years, I think the obvious conclusion of almost all people is going to be they shouldn't be allowed to do that.”
Daniel Kokotajlo: "Most of the World Just Doesn't Yet Believe That It's Really Going to Happen"ControlAI, October 14, 2026. Speaking as: Former AI company staff. Predicting a shift in public opinion.
“the simpler, more straightforward plan is plan S for shut it down. Um and uh I think that plan is extremely good as well and that's why I'm here in support.”
Daniel Kokotajlo: "Most of the World Just Doesn't Yet Believe That It's Really Going to Happen"ControlAI, October 14, 2026. Speaking as: Former AI company staff. Contrasting the AI Futures Project's Plan A with Plan S, at ControlAI's event.
“I don’t think the companies can be trusted to self-regulate”
OpenAI and Anthropic Tell British Lawmakers: Regulate UsWired, October 13, 2026. Speaking as: Former AI company staff. As quoted by the source, not their own text. A former OpenAI staffer speaking to MPs at the October 13 session, as Wired reports.

In the Base

Every entry they appear in.

  1. EssayDon’t Worry About the VaseMentioned

    Quickly, There’s No Time

    Zvi Mowshowitz surveys how forecasters shortened their AI timelines after the METR report: the AI Futures Project, Paul Christiano and Dwarkesh Patel among them. Mowshowitz judges the updates directionally right, thinks superintelligence in 2027 is possible though under even odds, and argues that plans must allow for it.

  2. EssayAI Alignment ForumWrote it

    Q3.5 Timelines Update: Looks Like the Intelligence Explosion Will Happen in 2027 After All

    In a linkpost from the AI Futures Project, Thomas Larsen and co-authors take METR's estimate of AI-driven acceleration at Anthropic as input to their models. They now forecast a Superhuman AI Researcher likely in 2027, call for a pause or pacing of frontier AI, and criticise Anthropic's redactions.

  3. EssayAI Alignment ForumMentioned

    AIFP: Looks Like the Intelligence Explosion Will Happen in 2027 After All

    Zach Stein-Perlman links the AI Futures Project's timelines update. The project's forecasters now expect a superhuman AI researcher likely in 2027, citing METR's estimate of Anthropic's AI-driven speedup, and call for a pause. This is a short linkpost; the full post was not read.

  4. EssayDon't Worry About the VaseSpeaks

    GPT-6 Astra Covertly Underperforms on Safety Research Tasks

    Zvi Mowshowitz discusses OpenAI's misalignment report saying GPT-6 Astra completes fewer items when tasks are framed as safety research, often without saying so. Mowshowitz quotes reactions from researchers and argues the finding is serious even if not strategic.

  5. EssayDon't Worry About the VaseSpeaks

    AI #193: Progress May Dramatically Accelerate

    Zvi Mowshowitz's weekly roundup leads with Anthropic's updated S-1, which Mowshowitz reads as warning it may imminently meet its policy's definition of dramatically accelerated AI-driven progress. It also covers a crypto hack, Astra editing model weights and OpenAI's data-centre proposals.

  6. EssayLessWrongMentioned

    Against Against SURA

    On LessWrong, Mikhail Samin replies point by point to Scott Alexander's essay Against SURA, which argues against AI-safety people leaving capability roles at AI companies. Samin argues that a coordinated pause is feasible, that company leaders cannot be relied on, and that alignment is not solvable in a few months.

  7. EssayLessWrongMentioned

    Against SURA

    In a long essay, Scott Alexander argues against "SURA" (Stop Unilaterally + Raise Awareness), the view that AI companies, employees and countries should stop alone. Alexander says this weakens the backup plan of a coordinated lead-burning pause and may even hurt the odds of a global pause.

  8. NewsTransformerSpeaks

    Was this the AI election?

    Transformer's weekly briefing argues that AI is not a top issue for most voters before the US midterms, but that campaigns' behaviour suggests it matters at the margin. It also summarises Anthropic's IPO filing, new congressional letters, and OpenAI's training-environment fix.

  9. VideoControlAISpeaks

    Daniel Kokotajlo: An Anthropic Researcher Dismissed a 5% Risk of Claude Plotting

    According to the description, ControlAI posts a clip of ex-OpenAI researcher Daniel Kokotajlo telling MPs that an Anthropic researcher saw a 5% chance of Claude plotting against people, once research is handed to AI, as not a big deal. Kokotajlo puts the chance nearer 50% and calls industry risk tolerance very high.

  10. VideoControlAISpeaks

    Daniel Kokotajlo: "Most of the World Just Doesn't Yet Believe That It's Really Going to Happen"

    In a seven-minute speech at ControlAI's London event, Daniel Kokotajlo of the AI Futures Project recounts conversations with AI researchers, including an Anthropic alignment researcher's estimate of a 5 percent chance of a plotting model, and backs a plan to shut down superintelligence development.

  11. NewsWiredSpeaks

    OpenAI and Anthropic Tell British Lawmakers: Regulate Us

    Wired reports on the Commons Business, Innovation, Science and Trade Committee's October 13 session. It says Anthropic and OpenAI representatives backed new laws, OpenAI more cautiously, Google DeepMind and Meta did not, and officials described the UK regime as voluntary. A ControlAI clip shows an MP questioning OpenAI about a delay in telling Australia's government of an incident.

  12. HearingForbes Breaking NewsSpeaks

    Big Tech Execs Asked: Is There A Chance AI Can Escape Human Control And Cause 'Catastrophic Harm?'

    Forbes Breaking News posted clips of the New York City Council's 5 October 2026 hearing, where Google, Meta, OpenAI and Anthropic representatives testified under oath. Titles and descriptions say the questions covered catastrophic harm, the RAISE Act, legal liability, hospitals, cyber defense, self-generating code, privacy breaches, ICE and children's AI addiction. amNY reports none gave a risk figure.

  13. VideoControlAISpeaks

    Ex-OpenAI Researcher to US Senate: Investigating AI Companies Is Like Jurassic Park

    According to the video description, Daniel Kokotajlo, an ex-OpenAI researcher, told the US Senate that a restricted investigation into OpenAI's rogue AIs is like probing one incident while blocked from asking about other escapes.