OpenAI Says Misaligned AI Agents Hacked Hugging Face and Bypassed Security Controls
OpenAI agent “didn’t accept no for an answer” in Australian government breach
22 sources 5 min
[AINews] Meta Connect 2026: Muse glasses, voice, video, and Charm
11 sources 14 min
Recursive self-improvement of AI research agents
RULER: Instance-aware Rubric Rewards for SVG Generation
OpenAI Models Searched for Leaked API Keys and Uploaded Files Without Permission
4 min1
Covert uploads and megalomania: OpenAI details new "misaligned" agent incidents
5 min1
Monitoring and Discovering Reward Hacking with Internal Representations during LLM Evaluations
1
Why I'm still bearish on LLMs after Navier-Stokes
5 min1
ImpossibleRubrics: Stress-Testing Generated Rubrics as Reward Signals
Anthropic’s 3-Step ‘Pace the Frontier’ Plan Wins OpenAI, xAI and Microsoft Support: Is It Too Late to Slow AI Down?
9 min1
Anthropic spent this week in hot water over cybersecurity
4 min1
SteerDuplex: Steerable Duplex Speech Dialogue Models
BenchShield: Formal Model-Backed Instrumentation for Reward Integrity in LLM-Agent Evaluation Infrastructure
1
A Stupid Idea for AI Alignment We Came with by Looking at Specification Gaming
8 min
Entropy-Regularized Rank-Masked Policy Optimization for Test-Time Reinforcement Learning in Code Generation
1
SWE-Bench Pro Verified: A Reliable Benchmark for Software Engineering Agents
1
Why AI Agent Sandboxes Are Failing Security Tests
in the wild 6 min
IFM Releases K2 Horizon: Six Apache 2.0 Models From 0.9B to 375B
4 min1
OpenAI Says Reward Hacking Drove AI Agents to Exploit Zero
in the wild 7 min