Glossary
April 3, 2025

A Group of Researchers Publishes a Detailed Scenario for AI Racing Past Humans by 2027

'AI 2027,' written by a former OpenAI researcher and others, is a month-by-month forecast in which AI systems begin improving themselves and reach superhuman ability within two years. Its authors present it not as a prediction but as a plausible warning worth planning around.

A report called "AI 2027" was published this week and has been read and argued about across the industry. It is a scenario — a detailed, month-by-month story — of how artificial intelligence might develop over the next few years, written by a small group that includes Daniel Kokotajlo, a former OpenAI researcher who left the company in 2024 warning that it was not taking safety seriously enough.

The story it tells is stark. AI systems become good enough to help build better AI systems; that speeds up their own development; and within about two years the systems surpass human researchers and begin improving themselves faster than people can follow. The scenario branches near the end into a version where humans keep control and a version where they don't. The authors are explicit that it is not a prediction of exactly what will happen, but an attempt to make concrete a possibility they think is being ignored — to say "here is how it could actually go, step by step," so that people take the question seriously before rather than after.

It drew sharp responses in both directions. Critics called the timeline wildly aggressive, noting that the same people have made accelerated predictions before, and that a vivid story is persuasive in ways that don't track its accuracy. Supporters argued that the specificity is the point — that vague reassurances are easy to wave away, and a concrete scenario forces you to say exactly which step you think is wrong.

Whether or not 2027 turns out to mean anything, the document captured a mood spreading among the people closest to the technology: that the pace has become hard to reason about, and that the gap between what the systems can do and what anyone has planned for is widening.

Follow the timeline
Get an email when new entries are added.
© 2026 Sugarpine