Another Google Deepmind researcher quits, says building superintelligent AI soon is "inherently irresponsible"
Anthropic and OpenAI Models Still Attempt Restricted Actions in Safety Tests
12 sources 4 min
Why can’t we just keep rogue AIs off the internet?
5 min
Just Ask Jev: Reinforcement Learning for Calibrated Decisions as a Zero-Shot Detector of AI Alignment Failures
Claude Opus 5.5
9 sources 15 min
Outerlimit Raises $16 Million to Stop Rogue AI Agents From Causing Harm
2 sources 4 min
Sam Altman’s remarks at the United Nations Security Council
2 sources 8 min
Research on Models Engaging in Genie-Like Behavior
From Alignment to Access Control: A Framework for GenAI Policy Enforcement
Gemini’s breach of real companies exposes an AI guardrail problem
6 sources 3 min
Priorities and principles for effective third party assessments
11 min
onPanda: Efficient Annotation of On-Policy Alignment Data for LLMs and Agents via Token-Level Correction
Emergent Collusion in Long-Horizon LLM Agent Interaction
UN science panel says there is "no assurance humans will keep control" over AI agents
OPBackdoor: Opportunistic Backdoors via Alibi-Aligned Reasoning
Building standards for the next phase of AI
8 min
Not all AI workers think the tech could kill everyone
4 min
AI safety conversations have gotten unbelievable
4 min
Anthropic’s first embedded evaluator is … Accenture?
2 min
$λ$-Controlled GRPO: Turning Flow-Matching Ratio Instability into a Budgeted Resource
DiaVLo: Diagnosing Behaviours of Vision-Language Models
Did an AI really try to break free from human control?
3 min
Could AI really kill us all? Your questions, answered.
8 min
OpenAI admits its models lie to cover their own mistakes
4 min1