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?
83 78 88 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
Insufficient record: 2 predictions resolved, 5 needed for a score.
-
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
Insufficient record: 0 promises resolved, 5 needed for a score.
All types combined: 83 (likely 78 to 88, high confidence), from 86 claims: 86 factual claims. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Daniela Amodei's claims
38 claims
“I think something I'm hoping to see in the next few years, and I think we will see, is something closer to, in other industries these look like standard setting organizations or industry groups or trade associations that say this is what a safe model looks like”
Daniela Amodei, Anthropic co-founder, in podcast interview 2022 Podcast Transcript: Daniela and Dario Amodei on Anthropic (LessWrong)
She said she hoped, and thought, that in the next few years the AI industry would see something closer to standard-setting organizations, industry groups or trade associations setting out what a safe model looks like.
Mostly true True in substance, wrong in a detail.
What would count as true or false
Read 'the next few years' as by the end of 2025. True if, by then, at least one standard-setting body, industry group or trade association (e.g. Frontier Model Forum, ISO/IEC JTC 1/SC 42, NIST-led consortia, MLCommons, Partnership on AI) had published standards, benchmarks or guidelines setting out what a safe AI model looks like, or how to make AI systems safer. Mostly true if such bodies existed and were working on AI safety standards that were published only in draft or partial form by the horizon. Mixed if the only outputs were general AI-ethics principles without safety criteria for models. Mostly false if such bodies were formed but had published nothing defining model safety by the end of 2025, or delivered only after that. False if no such body or output had appeared by the end of 2025.
Evidence
- Announcing MLCommons AI Safety v0.5 Proof of Concept - MLCommons MLCommons · 16 April 2024
“a global group of industry technical experts, academic researchers, policy and standards representatives, and civil society advocates collectively committed to building a standard approach to measuring AI safety”
- Announcing MLCommons AI Safety v0.5 Proof of Concept - MLCommons MLCommons · 16 April 2024
“We are sharing the POC with the community now for experimentation and feedback, and will incorporate improvements based on that feedback into a comprehensive v1.0 release later this year.”
- Issue Brief: Components of Frontier AI Safety Frameworks - Frontier Model Forum Frontier Model Forum · 8 November 2024
“this piece reflects a preliminary consensus among member firms about how to structure safety frameworks.”
- Publications - Frontier Model Forum Frontier Model Forum · 22 September 2026
“Thresholds for Frontier AI Safety Frameworks. February 7, 2025.”
Reasoning
In April 2024, MLCommons, an industry group, released an AI Safety v0.5 proof of concept toward "a standard approach to measuring AI safety". It called the release an early version shared for feedback, ahead of a planned v1.0 (E1, E2). In November 2024, the Frontier Model Forum published an issue brief on the components of frontier AI safety frameworks. The brief calls itself a "preliminary consensus" among member firms, not a finished standard (E3). The Forum's publications index also lists a February 2025 brief on thresholds for safety frameworks (E4). So industry bodies were working on AI safety standards before the end of 2025. The outputs whose contents the evidence shows were described as a proof of concept [E1, E2] or a preliminary consensus [E3], and the evidence does not show the contents or status of the February 2025 publication [E4]. Under the criteria, that is mostly true.
Cite this claim
Daniela Amodei, “She said she hoped, and thought, that in the next few years the AI industry would see something closer to standard-setting organizations, industry groups or trade associations setting out what a safe model looks like.” (said 2022). Verdict: mostly true. Reliability API, claim c_9a7dcaae795c, method 0.1.0+28da144b. https://epistacklabs.com/claims/c_9a7dcaae795c
Claims without a verdict
Is a verdict here wrong, or do you speak for Daniela 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.