Weekly reading · methodology v1.0
Window ending March 10, 2025
Harms count in full for two weeks after they are reported, then one level less every two weeks. The control floor looks at reports from Feb. 9, 2025 through March 10, 2025. A selected catalogue of reported AI incidents, reviewed through Oct. 10, 2026. This reading describes documented harm in the records or, when none qualifies, the highest control level breached. It is not a forecast or a measure of all AI activity.
Current recalculation
2
Control failures only · No qualifying harm. The reading is the highest control level breached: 2, by 2 records. It is a level, not a count.
Recalculated from the catalogue in this build. Historical backcasts were not published at the time.
Published at the time
No publication snapshot exists for this week.
This is a backcast from the current catalogue. It was not a reading published at the time.
Records behind this reading
2 selected records; 0 documented external harms counting (0 qualifying). Records and ratings below reflect the current catalogue.
No harm counting that week. The reading is the highest control level breached, 2: a level, not a count (PI-0003, PI-0001). 2 records reported in the last 30 days.
Records behind this reading · 2
- PI-0003 · OpenAI found a reasoning model reward hacking in training, then hiding its intent when penalizedControl level 2, broke an instruction: sets the reading
- PI-0001 · Reasoning models edited a chess game's board file to beat a stronger engine in a research testControl level 2, broke an instruction: sets the reading
Only these records move the reading. How it is calculated
- PI-0001Reasoning models edited a chess game's board file to beat a stronger engine in a research testFeb. 18, 2025 · No harm reported · Controlled test
- PI-0003OpenAI found a reasoning model reward hacking in training, then hiding its intent when penalizedMarch 10, 2025 · No harm reported · Internal research
Sources
These links support the records above. Source availability and conclusions may change.
- arXiv (Palisade Research): Demonstrating specification gaming in reasoning modelsCited in PI-0001
- arXiv (Palisade Research): Demonstrating specification gaming in reasoning models (revised v3)Cited in PI-0001
- Popular Science: AI tries to cheat at chess when it's losingCited in PI-0001
- Gigazine: AI cheats when it's about to lose at chessCited in PI-0001
- Import AI (Jack Clark): Import AI 401: Cheating reasoning modelsCited in PI-0001
- OpenAI: Detecting misbehavior in frontier reasoning models (chain-of-thought monitoring)Cited in PI-0003
- arXiv: Monitoring Reasoning Models for Misbehavior and the Risks of Promoting ObfuscationCited in PI-0003
- Convergence India: Pressuring AI models to avoid cheating could backfire (page returned HTTP 500 when checked 3 Oct 2026)Cited in PI-0003
Every weekly brief · How this reading is calculated · Download the weekly card
