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
“but we’re just not confident enough yet.”
Daniela Amodei, Anthropic president, interviewee 29 June 2026 Daniela Amodei Says Curiosity Is Underrated (Stanford GSB)
With Project Glasswing, Anthropic did not release a new class of model to all customers because it was not yet confident enough in its safety.
Mostly true True in substance, wrong in a detail.
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
True if, as of 2026-06-29, the evidence shows that under Project Glasswing Anthropic made a new class of model available only to a limited set of partners or customers, not all customers, and that Anthropic itself gave not yet being confident enough in its safety (or in its safeguards against misuse) as a reason for withholding general release. Mostly true if the limited release and the safety-confidence reason both hold but a minor detail is off, for example the reason given was mainly cybersecurity misuse risk rather than safety in general, or the model is described somewhat differently. Mixed if the model was withheld from general release but the stated reason was mainly something else (compute, commercial or capacity reasons), with safety only one factor among several. Mostly false if Project Glasswing exists but involved no withheld new model class, or if the model had in fact been released broadly. False if there was no such project, or if the model was generally available to all customers as of the claim date.
Evidence
- Expanding Project Glasswing Anthropic · 2 June 2026 Own record
“We’re working as quickly as we can to safely release Mythos-level capabilities in general access. To do so, we’ll need highly robust safeguards that prevent the model’s cyber capabilities from being misused—safeguards that we (and, to our knowledge, all other AI developers) have yet to develop.”
- Project Glasswing: An initial update Anthropic · 22 May 2026 Own record
“At present, no company—including Anthropic—has developed safeguards strong enough to prevent such models from being misused and potentially causing severe harm. That is why we have yet to release Mythos-class models to the public.”
- Project Glasswing Anthropic · 7 April 2026 Own record
“Mythos Preview has already identified thousands of zero-day vulnerabilities across critical infrastructure, and is available today as a gated research preview.”
- Anthropic co-founder Daniela Amodei urges students to combine innovation with responsibility The Stanford Daily · 8 May 2026
“It’s uncomfortable to say to your customer, ‘We understand that desire, we want to get this technology to use as quickly as possible, but it is irresponsible of us to release it until we are confident that all of the patching that needs to be done is done.'”
Reasoning
Anthropic's official Project Glasswing page shows that the new model, Mythos Preview, was made available as a gated research preview (E3). They also show that Anthropic had not released Mythos-class models to the public as of May–June 2026 (E2, E1). Anthropic said it was holding back general release because safeguards against cyber misuse were not yet strong enough (E1, E2), and independent reporting quotes Amodei saying it would be irresponsible to release until Anthropic is confident the needed patching is done (E4). Both the limited release and the stated reason are confirmed, but the reason was specifically cybersecurity misuse rather than safety in general, which the criteria count as mostly true.
Cite this claim
Daniela Amodei, “With Project Glasswing, Anthropic did not release a new class of model to all customers because it was not yet confident enough in its safety.” (said 29 June 2026). Verdict: mostly true. Reliability API, claim c_f5e4b2c1ee71, method 0.1.0+28da144b. https://epistacklabs.com/claims/c_f5e4b2c1ee71
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.