The Diary of a CEOChatGPT Whistleblower: They Offered Me $2m To Keep Quiet, No One Is Ready For What's Coming!
At a glance
WHAT IT’S REALLY ABOUT
AI forecaster warns superintelligence race risks catastrophe and power concentration
- Daniel Kokotajlo says the AI industry’s “open secret” is that superintelligence could arrive by the end of the decade and could plausibly lead to catastrophe, including scenarios as severe as human extinction.
- He describes two central worries: losing control of systems whose internal reasoning is hard to inspect, and the concentration of economic and political power if a small group controls superhuman AI.
- He claims leading labs are pursuing a strategy of automating their own R&D—starting with coding and moving toward fully autonomous AI research—creating race dynamics that compress timelines and amplify risk.
- He explains his resignation from OpenAI as driven by disillusionment, a desire to publish forecasting work publicly, and opposition to exit terms he says attempted to restrict criticism via an equity-linked clause.
- He contrasts a likely default “race” trajectory with a proposed alternative (AI 2040: Plan A) involving governmental slowdown, international verification, total research transparency, and mechanisms like a citizen’s dividend to manage job displacement and power shifts.
IDEAS WORTH REMEMBERING
5 ideasKokotajlo believes superintelligence is likely within years, not decades.
Daniel Kokotajlo argues that if AI becomes better than the best humans at everything and is paired with robotics, then society must prepare for extreme upside and extreme downside on very short timelines.
The biggest risks are loss of control and concentration of power.
He says the key danger is not just what AI can do, but that humans may be unable to verify whether increasingly powerful systems truly share human goals; he also flags heavy centralization if a few firms or leaders control “armies” of copies of the same model.
AI companies are trying to automate themselves to accelerate progress.
He describes major labs focusing first on automating coding and then the entire AI research loop so they can improve AI faster, which he says creates “intelligence explosion” dynamics and compresses timelines for widespread societal disruption.
He left OpenAI because he didn’t trust the industry to slow down when it matters.
Kokotajlo says his disillusionment grew as company narratives shifted from “we’ll pause near the edge to do safety” toward “it’s not that risky,” and he resigned to regain freedom to publish scenarios he thought the public needed to see.
He says OpenAI tied equity to silence, and he refused to sign.
He recounts refusing to sign exit paperwork containing a non-criticism clause tied to keeping equity, expecting to lose about $2 million; after public backlash and internal employee pushback, he says the company reversed course and let people keep equity.
WORDS WORTH SAVING
5 quotesThe scary open secret in the AI industry right now is that right now that is kind of just a hope.
— Daniel Kokotajlo
It's possible that we'll end up essentially creating a new species that ends up ruling the world instead of us.
— Daniel Kokotajlo
I think we are headed to a bad place if things don't change.
— Daniel Kokotajlo
If you wait until most people have lost their jobs to regulate the AI companies, that's already too late.
— Daniel Kokotajlo
It is pretty crazy to think that we're building a technology, a brain, that we don't understand.
— Steven Bartlett
High quality AI-generated summary created from speaker-labeled transcript.
