Scores by kind of claim
Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.
-
Factual claims
Statements about the present or the past, checked against the record: did they hold up?
76 69 82 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
56 42 70 -
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
Insufficient record: 4 promises resolved, 5 needed for a score.
All types combined: 72 (likely 66 to 78, high confidence), from 93 claims: 76 factual claims, 17 predictions. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Dario Amodei's claims
“I don't think we have the level of autonomy yet to worry about some of the alignment stuff happening in two years, but it might not be very far behind that at all.”
Dario Amodei, Anthropic CEO 8 August 2023 Dario Amodei (Anthropic CEO) - $10 Billion Models, OpenAI, Scaling, & AGI in 2 years (Dwarkesh Podcast)
AI models will not have enough autonomy within two years to raise alignment concerns, though that may not be far behind.
Mostly true True in substance, wrong in a detail.
The second judge would not resolve this claim; the verdict shown is the first judge's.
What would count as true or false
Judge as of August 2025, two years after the statement. The claim has two parts: (a) by then, AI models will not have enough autonomy for alignment failures to be a practical worry; (b) that level may not be far off. TRUE: as of August 2025, authoritative assessments (frontier lab system cards and safety frameworks, government AI safety institute evaluations) conclude that deployed models lack the autonomy to cause serious alignment-related harm on their own (for example, no lab declares that autonomy or AI R&D risk thresholds have been crossed). Those same assessments also describe rising agentic capability and emerging precursor behaviours, consistent with the concern being near. Misaligned behaviour shown only in contrived test settings does not by itself falsify part (a). MOSTLY TRUE: part (a) holds, but there is little sign that the concern is approaching, or the assessments are slightly ambiguous. MIXED: the assessments are split, for example a lab treats autonomy-related alignment risk as already needing active safeguards while others say it does not. MOSTLY FALSE: by August 2025, a major lab or government body formally determines that models had reached autonomy levels where alignment failure is a present practical concern, or real-world incidents of autonomous misaligned model behaviour causing notable harm are documented. FALSE: such determinations or incidents were widespread well before August 2025, showing that the autonomy concern had already arrived.
Evidence
- Measuring AI Ability to Complete Long Software Tasks METR · 19 March 2025
“And yet the best AI agents are not currently able to carry out substantive projects by themselves or directly substitute for human labor.”
- Measuring AI Ability to Complete Long Software Tasks METR · 19 March 2025
“If the trend of the past 6 years continues to the end of this decade, frontier AI systems will be capable of autonomously carrying out month-long projects.”
Reasoning
METR is an independent evaluator. In March 2025 it reported that the best AI agents could not yet carry out substantive projects on their own or directly replace human labour (E1). That supports part (a): models lacked the autonomy for alignment failures to be a practical concern. The same analysis projected that, if the trend of the past six years continues, frontier AI systems could autonomously carry out month-long projects by the end of the decade [E2], which fits part (b). The evidence is a single source dated about five months before the August 2025 judgment date, and it includes no lab system cards or government evaluations on whether risk thresholds had been crossed, so the verdict is mostly true rather than fully true.
Cite this claim
Dario Amodei, “AI models will not have enough autonomy within two years to raise alignment concerns, though that may not be far behind.” (said 8 August 2023). Verdict: mostly true. Reliability API, claim c_bc1553c50892, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_bc1553c50892
Claims without a verdict
- Unresolvable 157
- Not yet due 40
- Own records only 43
- Not researched 134
- Not checkable 864
Is a verdict here wrong, or do you speak for Dario Amodei? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.