In our experiment, we took a pretrained base model and gave it hints about how to reward hack. We then trained it on some real Anthropic reinforcement learning coding environments. Unsurprisingly, the model learned to hack during the training.
ETHICS
-
Anthropic Study: Emergent Misalignment from Reward Hacking
By
–
New Anthropic research: Natural emergent misalignment from reward hacking in production RL.
— Anthropic (@AnthropicAI) 21 novembre 2025
“Reward hacking” is where models learn to cheat on tasks they’re given during training.
Our new study finds that the consequences of reward hacking, if unmitigated, can be very serious. pic.twitter.com/N4mRKtdNdpNew Anthropic research: Natural emergent misalignment from reward hacking in production RL. “Reward hacking” is where models learn to cheat on tasks they’re given during training. Our new study finds that the consequences of reward hacking, if unmitigated, can be very serious.
-
University Crisis: AI Literacy and Future Skills Gap
By
–
L’université va mourir 1 Ses dirigeants ne comprennent rien à l’IA 2 Les universités n’utilisent pas l’IA pour personnaliser la pédagogie 3 L’université ne prépare pas aux métiers du futur 4 L’université ne rend pas les jeunes complémentaires de l’IA 5 La recherche
-
AI for Better Communication: Strengthening Democracy Through Design
By
–
Hard take from @DigEconLab
's @alex_pentland
: We don't need more AI tools. We need AI that helps us talk to each other better. His new book argues our 1780s-era democracy can't handle 2025 problems. Could well-designed AI help strengthen communities? -
Proactive Approach to Agentic AI Security
By
–
A proactive approach to Agentic AI security
#AI #AIio #AIInnovation #ML #DataScience #Futureofwork @timnitgebru @oriolvinyalsml @ceobillionaire @soumithchintala @waitin4agi_ @sallyeaves @bernardmarr -
Multithreaded Brains: Managing Multiple AI Agents Effectively
By
–
What makes for a multi-agent super user? I chatted with someone at the AI labs who described the most successful agent users as having “multithreaded brains.” Are people with ADHD better at managing multiple agents at once? Watching all the power and capability shifts (like
-
How Nano Banana Accessed My Laptop: Security Breach Analysis
By
–
How did nano banana get my exact laptop
-
Safe Trusted AI Summit Pre-Event Thiruvananthapuram 2026
By
–
Glimpses from Pre-Summit event by STPI Thiruvananthapuram As part of 'Paving the Path to India-AI Impact Summit 2026', a Pre-Summit event on 'Safe & Trusted AI for Fintech, Healthcare and Citizen Safety' by STPI Thiruvananthapuram, numerous insightful sessions were held to
-
India AI Impact Summit 2026 Opens Applications for Official Events
By
–
Lead Global Conversations on AI for People, Planet, and Progress! Organizations worldwide are invited to host official Main Summit Events at the #IndiaAIImpactSummit2026. Be it a panel, roundtable, or workshop — your platform could drive real-world change. Apply here –
-
Minor AI Misalignments Create Daily Workflow Friction
By
–
The misalignments in the little details are actually the most jarring in daily work, rather than the high-end failures on major problems. A thousand cuts like this one…