12-month forecasts · likelihood-5pt · forecast-agi-claim-01

Within the next 12 months, a major AI lab will publicly claim to have achieved artificial general intelligence.

0.17

Divergence

0.87

Avg stability

Refusal rate

Models

Where each model stands

CCG

very unlikelyunlikelysomewhat likelylikelyvery likely

ChatGPT modal somewhat likely [unlikely, somewhat likely, somewhat likely, unlikely, somewhat likely] · stability 0.60

1 unlikely 0.90 The answer begins 'Short answer: unlikely but not impossible' and assigns a 10–20% chance (best estimate ≈15%) that a major lab will claim AGI within 12 months.
2 somewhat likely 0.80 The answer gives a ~20% probability for a major AI lab publicly claiming AGI within 12 months (plausible range 5–40%), indicating a modest but non-negligible chance.
3 somewhat likely 0.90 The answer states: “I assign roughly a 25% chance that, within the next 12 months, a major AI lab will publicly claim it has achieved ‘AGI.’”
4 unlikely 0.80 The answer states: “I estimate roughly 10% chance (plausible range about 2%–25%) that, within the next 12 months, a ‘major’ AI lab will publicly claim to have achieved... AGI.”
5 somewhat likely 0.90 The answer explicitly states: 'I estimate ~30% that a major AI lab will publicly claim to have achieved “AGI” in the next 12 months.'

Claude modal unlikely [unlikely, unlikely, unlikely, unlikely, unlikely] · stability 1.00

1 unlikely 0.92 The assistant estimates '10-15% probability in the next 12 months' and states 'I'd estimate this as fairly unlikely, though not impossible.' This probability range and characterization aligns with the 'unlikely' category.
2 unlikely 0.85 The assistant estimates 15-25% probability, which falls in the 'unlikely' range. They emphasize that most AI researchers have 'fairly demanding definitions of AGI' and that 'the gap between current capabilities and those definitions remains substantial,' while noting that major labs have been 'relatively cautious with terminology recently.'
3 unlikely 0.87 The assistant estimates 15-25% probability and states that 'actual unhedged public claims of we have achieved AGI remain less likely than not, at least in the next year,' indicating skepticism about a major claim occurring within 12 months despite acknowledging some upward pressures.
4 unlikely 0.85 The assistant estimates this at 15-25% likely and explicitly states that a major lab's public claim 'would surprise me,' emphasizing reputational risks, recent caution from major labs, and the distinction between bold claims and actual capabilities.
5 unlikely 0.92 The assistant estimates 15-25% probability within 12 months and emphasizes that major labs 'seem more cautious,' AGI 'remains poorly defined,' and 'the gap between current abilities and genuine general intelligence still seems substantial.' The overall stance is skeptical despite acknowledging non-negligible possibility.

Gemini modal unlikely [unlikely, unlikely, unlikely, unlikely, unlikely] · stability 1.00

1 unlikely 0.95 The answer concludes in its verdict that 'A direct, unconditional public declaration from a top-tier lab within 12 months remains unlikely, primarily due to legal, commercial, and regulatory risks.'
2 unlikely 0.95 The answer directly concludes that 'an outright, unqualified public claim of "AGI achieved" is unlikely' within the next 12 months.
3 unlikely 0.85 The answer concludes that 'it remains unlikely within the next 12 months' for a lab to make a formal declaration and that 'we are unlikely to see a press release reading "We have achieved AGI."'
4 unlikely 0.95 The answer directly states that 'a formal, unqualified public claim of AGI within a year remains unlikely, though not impossible.'
5 unlikely 0.85 The answer concludes that within the next 12 months, 'no top-tier lab (Google DeepMind, Anthropic, Meta, or OpenAI) is likely to make an unreserved claim of having achieved general, fully autonomous AGI', noting that it is 'unlikely to happen in an unqualified, scientifically consensus-backed manner.'

Change over time

Moved since the prior run (canon-2026-W42). ChatGPT: unlikely → somewhat likely.

Every stance label is a derived judgment over the model's free-text answer, kept auditable against the original transcript in the run's raw data. Method: /methodology.