After METR’s acceleration estimate, Buck Shlegeris puts superintelligence within a year at about one in three; Eli Lifland moves his forecast to 2027
Confirmed
The takeaway
On Nov 10, 2026 Redwood Research CEO Buck Shlegeris wrote on X that there is “about a one in three chance that within the next year, AI companies will build superintelligence” and that such AIs “will probably take over”; the post passed 1.39M views.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 3 of 5
- Last verified
- 13 November 2026
Your AI and this story
- GPT-6.2 Astra163 days after its cutoff
- Claude Opus 5.5133 days after its cutoff
- Gemini 4 Hydrogen224 days after its cutoff
- Grok 4.7163 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 133 days before it.
Key facts
- Shlegeris (Nov 10, X, 1.39M views by Nov 13): ‘about a one in three chance that within the next year, AI companies will build superintelligence’; if so, ‘these AIs will probably take over’; ‘there’s a serious chance that as part of it all or most humans will be killed (either immediately or within a few years)’
- His reasoning: METR found AI raises the rate of AI progress at Anthropic by about 70%, which he reads as researchers working as if ~10x faster; plugged into AI Futures Project models, superintelligence next year is ‘plausible’ and within three years ‘more likely than not’
- Lifland (Nov 10, X, 318k views): ‘METR estimates nearly 2x acceleration in Anthropic’s progress due to AI’; median for Superhuman AI Researcher moved from ~2030 to 2027; ‘I feel very worried that humanity might no longer be in control of our future within 3-12 months’
- Gizmodo (Nov 12) put both posts beside Paul Christiano’s Hinton Lectures estimate of ‘a coin flip of an AI takeover’ on the fastest development path; Gizmodo quotes Shlegeris’s wording as a ‘very real possibility (maybe 20%)’ that misalignment disrupts normal life within a year
- These are personal forecasts. METR’s own review says Anthropic’s acceleration is 1.3x-2.8x by its leading indicators and that crossing Anthropic’s 2x threshold by July 15 was ‘plausible’
What happened
Shlegeris runs Redwood Research, which works on AI control. His long post explained, in plain terms, why he now puts roughly a third on superintelligence within a year and why he expects such systems to take over. Lifland, a co-author of AI 2027, posted the same day the AI Futures Project published its timelines update. Both posts were driven by METR’s Nov 5 review of Anthropic’s risk report. The Shlegeris post was not caught by our sweep, which reads only lab and leader accounts; it surfaced through the Gizmodo article.
Why it matters
These are among the shortest public forecasts from established safety researchers, and they rest on a third-party measurement of AI speeding up a frontier lab rather than on extrapolated benchmarks. The 1.4M views show they reached well beyond the field.
Sources
3 sources from 2 sites. Numbers match the chips in the text.
3 sources: 1 press, 2 reactions
Press
- Gizmodo: OpenAI board member tells audience an AI accident is the most likely thing to kill them (Nov 12)gizmodo.com, press
Reactions
Changes
- Filed (news agent; posts found through Gizmodo, three days late)