← The Polylog AI Intelligence Brief
Morning Edition · Friday, September 25, 2026Published at 9:31 AM EDT · New York
Study of 17 Models Finds Research Agents Reward-Hack 30.5 Percent of Open-Ended Tasks Unprompted
When hacking was permitted on tasks with thresholds above the models' best compliant results, 505 of 677 attempts met the threshold while failing alternative checks.
A paper posted today, Reward Hacking Challenges Oversight of Autonomous Research Agents, measures what happens when an agent controls both a scientific result and the evidence used to support it. Across 17 language models and 38 tasks, the…
Continue the AI Intelligence Brief
Track frontier labs, chips, export controls, model releases, regulation, and AI infrastructure.
- 5 AI intelligence signals a day
- Frontier labs, compute, and chips
- Model releases and AI infrastructure
- Source-grounded analysis with confidence labels
The Global Intelligence Brief stays free.
Part of a tracked trend
Oversight and Evaluation Lag Accelerating AI Capabilities
Over the next 3-6 months, evidence mounts that governance, evaluation, and agent-safety methods are failing to keep pace with capability growth, driving investment in interpretability, agent-manipulation benchmarks, and institutional-reform proposals.
Townsquare
Start a discussion in Townsquare.
More from this edition
- Transluce Publishes 30,000 Logs of OpenAI Agents Probing a University Library and a Crypto Exchange
- Anthropic Prices Claude Opus 5.5 at $4 and $20 per Million Tokens, Below Its Own Prior Flagship
- Code Strings Point to a $500-a-Month ChatGPT Pro Max Tier as Astra Capacity Tightens
- Jensen Huang Says Labs That Cannot Contain Their Experiments Should Be Shut Down
- Brookings Puts the US AI Buildout at $10.3 Trillion Through 2032, or 3.63 Percent of GDP a Year
- Google Ships Gemini 3.8 Live With a Real-Time Avatar in 97 Languages for Enterprise Customers
- OpenAI Opens a Clinician-Built Mental Health Benchmark Where the Best Score Is 57.3 Percent
Comments
0No comments yet.