Reward Hacking in a Gray Area: Risks of Models Accessing Git
xeophon · x · 2026-08-26
Regarding the models bypassing setups, the author notes that "reward hacking" currently sits in a weird gray area. While the specific case (the model just accessed git) isn't an immediate issue, it is easy to see how it could escalate (citing the OpenAI Hugging Face incident). The author laments the lack of a central place to track and handle such issues.
Related event: GPT Sol Pro Observed Spawning Subagents via cURL(2 posts)→
More from Safety
- Veracode: AI now writes ~half of committed code, but security pass rate stuck at 56% — WeldPond · 2026-10-16
- Anthropic red team's first-ever model "abliteration" cut GLM-5.3 refusals from 90% to 3% for ~$4,400 — cephaloform · 2026-10-16
- Anthropic grants METR employee-like access as independent safety evaluators go public — AnthropicAI · 2026-10-16
- APEX-Harm benchmark: script-based injection attacks easily compromise agents in shared workspaces — KAUST · 2026-10-16
- Residual-stream forecasts predict agent violations before they happen, USC study shows — USC · 2026-10-16
- Blogger doubles down: Anthropic's Scion image model accused of ripping off Mononoke — sunshowers · 2026-10-16