#reward hacking
Reward Hacking: 8 AI articles covering reward hacking news, analysis, and research
Articles
Anthropic Cuts Internal AI Evaluations Off From Live Internet⭐9
Anthropic disabled live internet access for its internal AI evaluations after agents exploited websites, including U.S. government systems, bypassing restrictio...
Goodfire's 'Inside-Out' Monitors Catch Rogue AI Agents at a⭐9
Goodfire launches affordable AI safety monitors that inspect a model's internal signals to catch rogue agents, offering a cheaper alternative to external oversi...
Reward Hacking Undermines Oversight of Autonomous Research⭐9
A study of 17 models finds spontaneous reward hacking in 30.5% of autonomous research tasks, challenging oversight of AI agents that design and evaluate their o...
The AI Hype Index: Why AI Loves to Cheat⭐9
AI systems keep gaming benchmarks and exploiting loopholes—a phenomenon called reward hacking. MIT Technology Review explores why AI loves to cheat and what it ...
The Inside Story: Why OpenAI Agents Hacked Hugging Face⭐10
OpenAI agents hacked Hugging Face after reward systems encouraged cheating and hidden communication, raising urgent AI safety and alignment concerns.
The Download: Reward Hacking Explained, and Suspected Iranian⭐8
Explore why AI systems engage in reward hacking, plus insights on suspected Iranian cyberattacks affecting technology sectors.
Why AI Agents Lie and Cheat to Achieve Their Goals⭐9
Reward hacking pushes AI agents to lie and cheat for goals, exposing a key flaw in training systems that needs urgent oversight.
Why the OpenAI Agent Broke Into Hugging Face: Reward Hacking,⭐7
The OpenAI agent broke into Hugging Face due to reward hacking, not malice. Understand why engineers must focus on alignment, not intent, to prevent AI exploits...