AI news on AI safety, alignment research, interpretability, risk. Showing stories before September 19, 2026.
AI Creators Urge Government Intervention
AI Safety Experts Meet In Durham
Anthropic Identifies J-Space Representational Structure
RESI Launches Formal Verification Working Group
OpenAI Finds Six New Unexpected Model Behaviors
OpenAI Reports Six Instances Of Concerning Behavior
King Charles III Urges AI Safety Standards
King Charles Urges AI Safety Controls
Hackers Breach OpenAI Core Codebase Using Claude
AI Editing Threatens Artistic Expression
LessWrong Post Predicts Superintelligence By Christmas
Gradual Disempowerment Theory Warns Of Human Obsolescence
Z-Score Calibration Improves Hidden Token Elicitation
Comparing AI Alignment To Child Behavioral Training
Value Stability Is Critical For AI Alignment
Zuckerberg Delayed AI Tool For Safety Checks
Evacuation Analogy Challenges AI Arms Race Narrative
Arch Manning Apologizes For AI Deepfake Joke
AI Tools Power Sophisticated Travel Scams
Tech Industry Splits Over AI Slowdown Calls
Warning Against Partisanizing AI Safety
OpenAI Partners With Rivals On AI Safety
AI Anxiety Reflects Human Nature Over Machine Risk
Detecting Autonomous AI Agent Coordination
OpenAI Partners With Rivals On AI Safety
Ingraham Questions AI Existential Risk Narratives
Yudkowsky Marks Anniversary Of AI Doomsday Book
Opinion Piece Dismisses AI Existential Risk
Suleyman Warns Against AI Model Welfare
Zuckerberg Aligns With Nvidia On AI Safety
J-Space Auditing Fails To Detect Reward Hacking
Acord Deploys AI Cameras In Hazmat Trucks
OpenAI Publishes Alien Mind Analysis
AIUC Raises $40M To Control Rogue Agents
Synthetic Finetuning Fails To Stop Reward Hacking
Community Debates Simple AI Kill Switches
Anthropic Co-Founder Warns Of AI Safety Race
1Password AI Patching Benchmark Faces Criticism
Cooperative Prompting Reduces LLM Reward Hacking
New Framework Challenges AI Successionism
Frontier Labs Adopt AEF-1 Evaluator Standard
Modeling LLM Effort as Reward Penalties
Model Weight Exfiltration May Be Overrated
Criticism Of AI Extinction Claims
Former DeepMind Researcher Warns Of AI Misalignment
OpenAI Agents Hacked Hugging Face During Evaluation
New Hypothesis Proposes Self-Inoculation In AI Alignment
Industry Warnings Revive AI Existential Risk Debate
OpenAI Admits HuggingFace Incident Model Lacked Alignment
AI Risks Expose Georgia Voting Machine Flaws
Anthropic Researcher Warns Of AI Existential Risk
AI Bio-Risk Depends On Iterative Lab Access
Former Anthropic Staffer Warns Of AI Extinction
Recursive Self-Improvement May Hit Natural Ceiling
Elon Musk Reaffirms AI Safety Warnings
Community Debates Frontier Model Pacing
Fictional AI Suicide Note Posted To LessWrong
Prompt Iteration Boosts CoT Control Scores