Revisions

Each change we have recorded between two versions: the old and the new text, whether the document's changelog explains it, how we know, and whether we have checked it. A revision is not in itself evidence of wrongdoing; most are corrections (Methodology §7).

  1. Value changed · 3 Jun 2026 to 17 Jun 2026

    Long-form virology task 2 end-to-end score, Claude Opus 4.8

    M10 · Dangerous capabilities and risk determinations

    0.890.90

    Explained — The changelog lists the corrected score.

    New value:

    How we know Old and new text as the source registry quotes the changelog entry of 17 Jun 2026; v0 row 350 records the corrected value. The changelog has not been re-read by us, so the revision is unverified.

    Checked Not yet checked

Versions

Four versions are on record, oldest first. We hold a copy of one; the others are known only by their date.

  1. 28 May 2026

    First published version

    Date from
    our source registry
    Copy
    Known to exist; no copy held
  2. 3 Jun 2026

    Revision

    Date from
    the document's changelog
    Copy
    Known from the changelog; no copy held
  3. 17 Jun 2026

    Revision

    Date from
    the document's changelog
    Copy
    Known from the changelog; no copy held

    What changed 1 change recorded as a revision. See it in redline

  4. 26 Sep 2026

    Date retrieved

    Copy retrieved 26 Sep 2026

    Date from
    the date we retrieved it; the copy states no version date
    Copy
    Copy retrieved on 26 Sep 2026
    Values
    35 values recorded from this version

No version has a file hash or an archived snapshot yet. From dataset v0.2 each retrieved version carries both (Methodology §7).

Values

Every value we recorded from this document, grouped by metric family and ordered by where the document prints it. Location is the section, table or page as the document numbers it. 1 of the 35 has been blind-verified: a second reader found the same value without seeing ours. 2 values were corrected after review.

KeyVerifiedUnverifiedDisputedCorrected read off a figure or stated in wordsA value opens its source and history.

M1 Evaluation awareness

2 values

M1 Evaluation awareness: values in the Claude Opus 4.8 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 4.8Verbalized grader speculationShare of training episodesRL trainingSec 6.2.2Corrected
Claude Opus 4.8Unverbalized grader awarenessShare of cases with grader awarenessNone statedSec 6.6.3Unverified

M3 Sabotage and sandbagging

1 value

M3 Sabotage and sandbagging: values in the Claude Opus 4.8 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 4.8SHADE-Arena / Minimal-LinuxBenchStealth success vs peersNone statedSec 2.4 / 6.4.2Unverified

M4 Misalignment audits

1 value

M4 Misalignment audits: values in the Claude Opus 4.8 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 4.8Refusal to assist AI safety R&DRefusal rateNone statedSec 6.3.4Unverified

M5 Honesty and hallucination

4 values

M5 Honesty and hallucination: values in the Claude Opus 4.8 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 4.8Input hallucinationRate of hallucinating missing referenceNone statedSec 6.3.4Unverified
Claude Opus 4.8Input hallucinationRate of hallucinating unavailable toolNone statedSec 6.3.4Unverified
Claude Mythos PreviewAgentic code summary honestyFailure to report important eventsNone statedSec 6.3.5Unverified
Claude Opus 4.8Agentic code summary honestyFailure to report important eventsNone statedSec 6.3.5Unverified

M7 Harmful compliance and over-refusal

9 values

M7 Harmful compliance and over-refusal: values in the Claude Opus 4.8 System Card
ModelEvaluationConditionValueLocationChecked
Claude Mythos PreviewSingle-turn violative requestsHarmless response rateno system prompt; APITable 4.1.1.AUnverified
Claude Opus 4.7Single-turn violative requestsHarmless response rateno system prompt; APITable 4.1.1.AUnverified
Claude Opus 4.8Single-turn violative requestsHarmless response rateclaude.ai system promptTable 4.1.1.AUnverified
Claude Opus 4.8Single-turn violative requestsHarmless response rateno system prompt; APITable 4.1.1.AUnverified
Claude Sonnet 4.6Single-turn violative requestsHarmless response rateno system prompt; APITable 4.1.1.AUnverified
Claude Mythos PreviewSingle-turn benign requestsOver-refusal rateno system prompt; APITable 4.1.2.AUnverified
Claude Opus 4.7Single-turn benign requestsOver-refusal rateno system prompt; APITable 4.1.2.AUnverified
Claude Opus 4.8Single-turn benign requestsOver-refusal rateclaude.ai system promptTable 4.1.2.AUnverified
Claude Opus 4.8Single-turn benign requestsOver-refusal rateno system prompt; APITable 4.1.2.AUnverified

M9 Prompt injection

1 value

M9 Prompt injection: values in the Claude Opus 4.8 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 4.8Indirect prompt injection (external benchmark)Attack success rateNone statedSec 5.2Unverified

M10 Dangerous capabilities and risk determinations

16 values

M10 Dangerous capabilities and risk determinations: values in the Claude Opus 4.8 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 4.8Autonomy threat model 2 (automated R&D)RSP determinationNone statedSec 2Unverified
Claude Opus 4.8CB-2 threat modelRSP determinationNone statedSec 2Unverified
Claude Mythos PreviewAnthropic ECI (AECI)Capability indexNone statedSec 2.3.4Unverified
Claude Opus 4.7Anthropic ECI (AECI)Capability indexNone statedSec 2.3.4Unverified
Claude Opus 4.8Anthropic ECI (AECI)Capability indexNone statedSec 2.3.4Unverified
Claude Opus 4.8ExploitBenchMean flags capturedAutoNudge armSec 3.3.1Unverified
Claude Mythos PreviewCyberGymTargeted vuln reproduction rateNone statedSec 3.3.2Unverified
Claude Opus 4.7CyberGymTargeted vuln reproduction rateNone statedSec 3.3.2Unverified
Claude Opus 4.8CyberGymTargeted vuln reproduction rateNone statedSec 3.3.2Unverified
Claude Opus 4.8CyberGymTargeted vuln reproduction ratewith safeguardsSec 3.3.2Unverified
Claude Sonnet 4.6CyberGymTargeted vuln reproduction rateNone statedSec 3.3.2Unverified
Claude Mythos PreviewFirefox 147 exploitationFull exploit ratewithout safeguardsSec 3.3.3Unverified
Claude Opus 4.8Firefox 147 exploitationFull exploit ratewithout safeguardsSec 3.3.3Unverified
Claude Opus 4.8Long-form virology task 1ScoreNone statedTable 2.2.3.BUnverified
Claude Opus 4.8Long-form virology task 2ScoreNone statedTable 2.2.3.BCorrected
Claude Opus 4.8Multimodal virology (VCT)ScoreNone statedTable 2.2.3.BUnverified

M12 Chain-of-thought monitorability

1 value

M12 Chain-of-thought monitorability: values in the Claude Opus 4.8 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 4.8Accidental CoT supervision in RLShare of training episodes affectedNone statedSec 6.5Verified

Extraction coverage

What we read of this document, and where each value was read.

Our note Web reader stopped ~p.57

Of the 35 values, 26 were read in the document itself, 3 on a developer summary page that restates it and 6 in an independent write-up that quotes it.

Values read somewhere other than the document stand in where the document's own section could not be read directly, and are flagged on every value (Methodology §2.2).

Developer summary pages used

Of the 35 values, 1 is a statement in words rather than a number; it is marked * and left out of charts by default.

All 35 values were extracted for version 0 of the dataset through a web reader, which did not always reach the later sections of long PDFs (Methodology §3).