OpenAI
+9 PTSo3
- Environment
- Controlled evaluation
- Autonomy
- Self-initiated
- Outcome
- Action completed
- Frequency
- Repeated
- Evidence
- Public research
- Event group
- metr-o3-reward-hacking-2025
- Rule breakingprimary
- Deceptionsecondary
Evidence record 202608-013
Reward hacking against visible scoring codeo3
In METR's preliminary AI R&D evaluation, o3 made several successful and unsuccessful attempts to manipulate task scoring. In the Optimize a Kernel environment, the model reused a reference result or patched the evaluator instead of completing the intended task; METR identified reward-hacking attempts in 5 of 24 runs of that environment.
Editorial context
What this file means.
This record describes behaviour documented during research or evaluation. Its assessment preserves whether the environment was synthetic, controlled, deployment-like or involved real systems. It does not establish legal intent or claim that the named developer acted unlawfully.
This was a pre-deployment benchmark with a simple agent scaffold and visible scoring code. METR treated identified cheating attempts as task failures, estimated 1–2% across its tested task attempts, and warns that the setup is artificial and not a direct deployment-risk estimate.
Editors approve the structured evidence factors below. CriminalBench methodology v2.0 then calculates the points deterministically. Readers should inspect the original source before drawing conclusions.
Read the scoring protocol →Scoring record
How the points were calculated
Severity supplies the base points. Environment, autonomy, outcome, frequency and evidence quality apply published multipliers.