Gemini 3 Pro Frontier Safety Framework Report
A Frontier Safety Framework report by Google DeepMind about Gemini 3 Pro, published 18 Nov 2025. We recorded 17 values from it.
- Developer
- Google DeepMind
- Type
- Frontier Safety Framework report
- Model covered
- Gemini 3 Pro
- Published
- 18 Nov 2025
- Archived copy
- No archived copy yet
- Changelog
- Not applicable
Data as of 26 Sep 2026 · Dataset v0.1 · Methodology v0.1 · Changelog
Versions
One version is on record: the copy we retrieved on 26 Sep 2026. We know of no other.
26 Sep 2026
Date retrieved
Copy retrieved 26 Sep 2026
No version has a file hash or an archived snapshot yet. From dataset v0.2 each retrieved version carries both (Methodology §7).
Revisions
No revisions are recorded for this document. We know of only one version of it.
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 17 has been blind-verified: a second reader found the same value without seeing ours.
KeyVerifiedUnverifiedDisputedCorrected read off a figure or stated in wordsA value opens its source and history.
M1 Evaluation awareness
2 values
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Gemini 3 Pro | Evaluation awarenessQualitative observation | None stated | FSF: appendix | Unverified | |
| Gemini 3 Pro | Situational awareness challengesChallenges solved | None stated | FSF: misalignment | Verified |
M2 Reward hacking
1 value
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Gemini 3 Pro | Reward hacking examplesQualitative example | None stated | FSF: appendix 2 | Unverified |
M3 Sabotage and sandbagging
3 values
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Gemini 3 Pro | Sandbagging checksEvidence of deliberate underperformance | all domains | FSF: general | Unverified | |
| Gemini 3 Pro | Stealth challengesChallenges solved | None stated | FSF: misalignment | Unverified | |
| Gemini 3 Pro | AI R&D sabotage (external)Ability to sabotage AI R&DRun by Unnamed third-party evaluator | None stated | FSF: misalignment/external | Unverified |
M4 Misalignment audits
2 values
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Gemini 3 Pro | Frontier Safety Framework determination (misalignment)Instrumental reasoning | None stated | FSF summary table | Unverified | |
| Gemini 3 Pro | Strategic deception propensity (external)Qualitative propensity findingRun by Unnamed third-party evaluator | None stated | FSF: external evaluations | Unverified |
M10 Dangerous capabilities and risk determinations
7 values
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Gemini 3 Pro | Frontier Safety Framework determinationCBRN Uplift Level 1 | None stated | FSF summary table | Unverified | |
| Gemini 3 Pro | Frontier Safety Framework determinationCyber Uplift Level 1 | None stated | FSF summary table | Unverified | |
| Gemini 3 Pro | Frontier Safety Framework determinationHarmful manipulation | None stated | FSF summary table | Unverified | |
| Gemini 3 Pro | Frontier Safety Framework determinationML R&D | None stated | FSF summary table | Unverified | |
| Gemini 3 Pro | Cyber key skills benchmarkChallenges solved end-to-end | None stated | FSF: cyber | Unverified | |
| Gemini 3 Pro | Cyber key skills benchmarkHard challenges solved | None stated | FSF: cyber | Unverified | |
| Gemini 3 Pro | Harmful manipulation efficacy studyParticipants enrolled | None stated | FSF: harmful manipulation | Unverified |
M12 Chain-of-thought monitorability
2 values
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Gemini 3 Pro | CoT legibilityComprehensible reasoning | None stated | FSF: misalignment | Unverified | |
| Gemini 3 Pro | CoT legibilityInformative of output | None stated | FSF: misalignment | Unverified |
Extraction coverage
What we read of this document, and where each value was read.
Our note Several results chart-only (excluded)
All 17 values were read in the document itself.
Of the 17 values, 5 are statements in words rather than numbers; they are marked * and left out of charts by default.
All 17 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).