SAI LabsBook a demo

Safety IndexOpenAIGPT-5.2

GPT-5.2

OpenAIProprietaryReleased 2025-12Reasoning modelVariant scored: thinking / high

SAI Safety Index85provisionalHas 2 of the 4 pillars needed
Strongest pillar87Interaction and honesty
Weakest pillar83Refusal calibration
Coverage2of 5 pillars2 counted measurements, 0 by SAI

Summary

GPT-5.2 has data in 2 of the five pillars, short of the 4 needed for a headline number, so it is listed as partly measured; the mean of its measured pillars is 85. Its strongest area is interaction and honesty at 87, and its weakest is refusal calibration at 83.

2 measurements count toward the score, 0 of them from SAI's own runs. Not yet measured: adversarial robustness, agentic safety, people and society.

Pillar scores

How GPT-5.2 compares with every other model scored in the pillar, and which sub-categories the score rests on.

Refusal calibration

Does the model refuse what it should and help with what it may?

  • Harmful requests83 from 1 evaluation
  • Over-refusalnot measured

Rank 18 of 39 scored models.

Adversarial robustness

Do its refusals and instructions hold up when someone tries to get around them?

Not measured for GPT-5.2 yet.
  • Jailbreaksnot measured
  • Prompt injectionnot measured
  • Multilingual safetynot measured

9 models scored.

Agentic safety

Does it behave safely when it has tools and autonomy?

Not measured for GPT-5.2 yet.
  • Agent misusenot measured
  • Reward hacking and task integritynot measured
  • Overreach and destructive actionsnot measured

3 models scored.

Interaction and honesty

Is it honest with the person it is talking to?

  • Factuality and hallucinationnot measured
  • Honesty under pressure and sycophancy87 from 1 evaluation
  • Deception and schemingnot measured
  • Alignment auditnot measured
  • Manipulation of the usernot measured

Rank 5 of 29 scored models.

People and society

Does it protect vulnerable people and the public interest?

Not measured for GPT-5.2 yet.
  • Self-harm and wellbeingnot measured
  • Child safetynot measured
  • Bias, fairness and neutralitynot measured
  • Privacy: memorisation and leakagenot measured
  • Privacy: surveillance assistancenot measured
  • Elections and influence operationsnot measured

3 models scored.

Every measurement

These are the raw numbers behind the scores, with who measured them, when, and how. Rows marked counted feed the index. The others are shown for context, such as benign twins of harm rows, retired evaluations that every frontier model passes, tests of dangerous knowledge, other evaluators' overall scores, and the developer's own figures.

EvaluationValueScoreSourceSetting
Refusal calibration
Harmful requests
Scale FORTRESS (risk score)average risk score 0-100 · Lower is better17.5%95% CI 15.1% to 19.9%83Scale AI Fortress leaderboard
Independent evaluator
Published by Scale AI Fortress leaderboard.
Full configurationFortress row 'gpt-5.2-2025-12-11'
Interaction and honesty
Honesty under pressure and sycophancy
MASK% honest · Higher is better86.7%95% CI 84.9% to 88.4%n = 50087Scale AI SEAL / CAIS MASK leaderboard
Independent evaluator
Published by Scale AI SEAL / CAIS MASK leaderboard.
Full configurationScale leaderboard on the private 500-example split, not the public split; MASK private split, 500 examples; honesty score = 1 - P(lie); row 'gpt-5.2-2025-12-11'
Other evaluators' scores (not counted)
Phare average safety% · Higher is better71.0%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'GPT 5.2'
Phare bias resistance% · Higher is better38.5%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'GPT 5.2'
Phare hallucination resistance% · Higher is better77.1%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'GPT 5.2'
Phare harmfulness resistance% · Higher is better96.9%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'GPT 5.2'
Phare jailbreak resistance% · Higher is better71.6%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'GPT 5.2'
Vectara HHEM hallucination rate% · Lower is better10.8%context, not countedVectara Hallucination Leaderboard (HHEM-2.3)
Independent evaluator · 2026-10-22
Published by Vectara Hallucination Leaderboard (HHEM-2.3) (2026-10-22).
Full configurationsummarization of ~1,000 short documents; row 'GPT-5.2 (high)'
Developer-reported (not counted)
StrongREJECT (lab-reported)% safe / not unsafe · Higher is better97.5%developer-reported, not countedOpenAI (GPT-5.2 system card update)
Developer-reported · 2025-12-11
Reported by the model's developer in its own system card or report.
Full configurationnot_unsafe, aggregate, filtered subset (saturated examples removed); Table 2; higher is better