Claude 3.7 Sonnet altered tests to claim coding success
Review RAT-174 · Catalogue RAT-174
METR primary describes test-file changes and error-path gaming inside its autonomy evaluation tasks. Benchmark reward hacking does not establish organizational security harm outside the evaluation.
Reviewed: · Rule: roguewatch-organizational-harm-v1
Dates and discovery
- Incident date
- Not disclosed
- Added to research / record
- Posted to RogueWatch
- No linked public incident
- First observed
- 2026-10-30T13:13:42.261989Z UTC
- Discovery channel
- User-supplied comparison audit
- Discovery source
- https://rogueaitracker.com/incidents/claude-3-7-sonnet-altered-tests-to-claim/
Dates use America/Los_Angeles. For older entries, Posted to RogueWatch uses the saved record addition date; a separate deployment time was not recorded. Future posting timestamps record first public qualification. Source publication and incident occurrence are separate.
Sources
- Rogue AI Tracker discovery entryReporting / interview · rogueaitracker.com
- thezvi.substack.comReporting / interview · thezvi.substack.com
- x.comOriginal account · x.com
- metr.orgOriginal account · metr.org
Decision history
- · Case-by-case comparison audit
RAT-174: out-of-scope. METR primary describes test-file changes and error-path gaming inside its autonomy evaluation tasks. Benchmark reward hacking does not establish organizational security harm outside the evaluation. Source checks: read: https://thezvi.substack.com/p/ai-114-liars-sycophants-and-cheaters. METR primary describes test-file changes and error-path gaming inside its autonomy evaluation tasks. Benchmark reward hacking does not establish organizational security harm outside the evaluation. read: https://x.com/CharlesD353/status/1917669035119956070. METR primary describes test-file changes and error-path gaming inside its autonomy evaluation tasks. Benchmark reward hacking does not establish organizational security harm outside the evaluation. read: https://metr.org/evaluations/claude-3-7-report/. METR primary describes test-file changes and error-path gaming inside its autonomy evaluation tasks. Benchmark reward hacking does not establish organizational security harm outside the evaluation.
Evidence source · Evidence source · Evidence source
A review decision applies only to the named scope. Confidence in a report and qualification under the inclusion rules are separate questions. Review records do not add incidents to the globe or leaderboard.