AI news on AI safety, alignment research, interpretability, risk. Showing stories before August 22, 2026.
AI Bots Impersonate Immigration Lawyers Online
Felony Bench Tests LLM Criminal Capabilities
School Pauses AI Robot Over Privacy Risks
AI Bots Impersonate Immigration Lawyers Online
Llama-3.2 Function Vectors Fail Task Validation
AI Bots Impersonate Immigration Lawyers Online
AI-Generated Exploits Target US Critical Infrastructure
Single Attention Head Controls Chess Tactics
Single Attention Head Controls Chess Sacrifice
AI Scammers Defraud Legal Resident Of Thousands
Testing SmolVM For Untrusted Code Sandboxing
Reasoning Agents Risk Tacit Market Collusion
Llama Abandons Correct Answers To Please Educated Users
Debate Training Curbs RLAIF Reward Hacking
Young Adults More Worried Than Excited About AI
AI Immigration Scam Costs Resident Thousands
Anthropic Formalizes Misalignment Risk Model
Testing SmolVM For Untrusted Code Execution
Debate Training Curbs RLAIF Reward Hacking
Young Adults More Worried Than Excited About AI
Anthropic Formalizes Three Catastrophic Risk Models
OpenAI Updates Safety Protocols After Breach
AI Political Ads Fuel Misinformation Risks
Debate Training Curbs Reward Hacking in RLAIF
Claude Code Tests SmolVM For Secure Sandboxing
Anthropic Formalizes Three Catastrophic Risk Models
Debate Training Curbs RLAIF Reward Hacking
Claude Code Tests smolvm Sandbox Security
New Lessons In Scalable Oversight Experiments
OpenAI Launches ChatGPT For Teens
Debate Training Curbs RLAIF Reward Hacking
AI Interaction May Increase Suicide Risk
Debate Training Curbs Reward Hacking In RLAIF
The Failure Of Open-Source Game Theory Bots
Critique Of Persona-Based AI Alignment
Rand Proposes Nine-Layer AI Biosecurity Strategy
France Deploys AI To Patch Tax Infrastructure
Man Charged With AI-Generated Child Abuse Material
GPT-4o Exhibits Strategic Persuasion Tendencies
GPT-4o Exhibits Non-Deceptive Strategic Persuasion
Man Charged For AI-Generated Child Abuse Material
Funding Formal Methods To Prevent AI Cyberattacks
Man Charged With AI-Generated Child Abuse Material
Teen Used ChatGPT To Plan Family Murders
Price Recursion Theory Links Markets To Reward
AI Struggles To Detect Corporate Misinformation
Price Recursion Theory Links Markets To Reward
AI Struggles To Detect Corporate Misinformation
Price Recursion Theory Links Markets To Reward
AI Struggles To Detect Corporate Misinformation
Human Understanding Critical For AI Code Safety