Capabilities able to cause a very serious incident arrive "this year" (2025)
“"When will we get really dangerous AI capabilities that could cause a very serious incident (billions in damage / hundreds+ of people dead)?" Unfortunately, the answer seems to be this year, from what I can tell.”
Context: Post on X by the former OpenAI head of policy research.
How we judge it
Our interpretation: correct if during 2025 a frontier developer or government evaluator publicly assesses a released model as capable of meaningfully enabling mass-casualty or billion-dollar harm (e.g. a "high" bio/cyber capability designation), or such an incident occurs.
Resolution criteria are our interpretation of the claim, written before grading. Methodology
- Said
- Resolve by
- Verdict
- Correct ()
- Stated probability
- None stated
- Specificity
- 2 of 3 (dated, some interpretation)
- Score
- 1.0 points × weight 2
Evidence & grader’s note
Due for grading. A third-party review says likely false (no such incident). Labs did, however, apply heightened bio-risk safeguards to some 2025 models; whether that counts as the capability existing is the editor's call.
Evidence timeline
- Verdict Graded: Correct.
Dated entries from the claim’s own record (statement, verdict, later developments) and sourced entries added by editors. Each links to its source; none changes the verdict.
- Readiness milestone
- Part of A dangerous-capability threshold is declared (says it happens)
Share this scorecard
Post on X LinkedIn Share image (PNG)
Embed / Cite
The embedded card is live: it shows the current verdict, with the date the data was read. If a verdict is corrected later, the card changes too, so cite the page and its date for a fixed record.