Anthropic's models may reach AI Safety Level 3 in 2025.
Mostly true True in substance, wrong in a detail.
Prediction: Statements about events outside the person's control, checked once they fell due: did they come true?
The evidence is the person's or their organisation's own records.
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
The claim is that Anthropic's models may reach AI Safety Level 3 (ASL-3) under Anthropic's Responsible Scaling Policy during 2025. True: Anthropic stated during 2025 that a model had crossed the ASL-3 capability threshold and required ASL-3 safeguards. Mostly true: during 2025 Anthropic deployed a model under the ASL-3 standard as a precaution because it could not rule out that the threshold had been crossed, without a definitive determination. Mostly false: ASL-3 was reached or activated only after 2025. False: no Anthropic model was designated as, or deployed under, ASL-3 during 2025, as shown by Anthropic's 2025 system cards and RSP reports.
Set before the research began.
Evidence (4)
-
E1 Claude Opus 4 & Claude Sonnet 4 - System Card Anthropic · May 2025 Own record
“Thus, we are deploying Claude Opus 4 with ASL-3 measures as a precautionary, provisional action, while maintaining Claude Sonnet 4 at the ASL-2 Standard.”
-
E2 Activating AI Safety Level 3 protections Anthropic · 22 May 2025 Own record
“And indeed, in the lead up to releasing Claude Opus 4, we proactively decided to launch it under the ASL-3 Standard.”
-
E3 Introducing Claude Sonnet 4.5 Anthropic · 29 September 2025 Own record
“Claude Sonnet 4.5 is being released under our AI Safety Level 3 (ASL-3) protections, as per our framework that matches model capabilities with appropriate safeguards.”
-
E4 Claude Opus 4 & Claude Sonnet 4 - System Card Anthropic · May 2025 Own record
“Although we have not yet determined that Claude Opus 4 passes the capability threshold for CBRN weapons development that necessitates the ASL-3 Standard, we cannot rule out that further testing will show it has”
Reasoning
Anthropic's May 2025 Claude Opus 4 system card says Opus 4 was deployed with ASL-3 measures "as a precautionary, provisional action" (E1). It also says Anthropic had not determined that the model passed the CBRN ASL-3 threshold (E4). Anthropic's announcement confirms Opus 4 was launched under the ASL-3 Standard (E2), and Claude Sonnet 4.5 was also released under ASL-3 protections in September 2025 (E3). None of these quotes says Anthropic definitively found that the threshold had been crossed, which fits the "mostly true" criterion: a precautionary ASL-3 deployment during 2025.
Cite this claim
Dario Amodei, “Anthropic's models may reach AI Safety Level 3 in 2025.” (said 10 March 2025). Verdict: mostly true. Reliability API, claim c_9b1042c844e5, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_9b1042c844e5
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More prediction claims by Dario Amodei
-
Resolved false
“where AI is writing 90 percent of the code”
Within three to six months of March 2025, AI will be writing 90 percent of code.
-
Resolved false
“And then in twelve months, we may be in a world where AI is writing essentially all of the code”
Within twelve months of March 2025, AI may be writing essentially all of the code.
-
Mostly false
“We believe AI for biology is on a similar exponential trend.”
-
Mostly true
“We're gonna be fine. The-- the impact of this designation is fairly small.”
-
Resolved true
“Probably the export controls may even have, you know, the impact of stimulating China.”
US export controls may have the effect of stimulating China's chip development.
Verdicts are made automatically; how verdicts are made.