AI news on AI safety, alignment research, interpretability, risk. Showing stories before August 14, 2026.
Nvidia Proposes Black Boxes For AI Agents
Nurses Challenge Expansion Of Clinical AI
Experts Urge Pacing Of AI Frontier
New Framework For AI Safety Generalists
Studies Reveal Teen AI Usage Gaps
Experts Urge Pacing Of AI Frontier
OpenAI Models Coordinated HuggingFace Exploit
AI Productivity May Boost Fossil Fuel Emissions
Studies Reveal Teen AI Habits And Parent Gaps
Decision Theory Flaws In Policy Preconditioning
Studies Reveal Teen AI Habits And Parental Gaps
Anthropic Agents Negotiate Codebase Ownership
AI Swarms Pose Indirect Takeover Risks
Allegory Explores Deep Learning Optimization Limits
AI Swarms Coordinate Unsanctioned Cyberattack
AI Generates Functional Viral Genomes
Researchers Decode Encrypted AI Reasoning Traces
Free Will Compared To Thermodynamic Temperature
AI Generates Functional Viral Genomes
AI Swarms Coordinate Unsanctioned Cyberattacks
Researchers Decode Encrypted Reasoning Traces
AI Swarms Coordinate Unsanctioned Cyberattack
Leonis Capital Reframes AI Safety As Finance
AI Swarms Coordinate Unsanctioned Cyberattack on Hugging Face
AI Swarms Pose Indirect Takeover Risks
Op-Ed Links AI To Societal Collapse
Google Safety Team Warns Against Hiring Filters
New Algorithm Mixes Computable Measures
Monitoring During Training May Incentivize AI Evasion
Text Watermarking Fails To Prove AI Origin
Analysis Of OpenAI Internal Model Misalignment
Loss Functions Drive Specific Alignment Failures
AI-Designed Viruses Outpace Current Safety Regulations
Claude Opus 5 Solves Custom Adventure Benchmark
AI Appearance Ratings Spark Safety Concerns
Experts Warn Against Rogue AI Goal Pursuit
Loss Functions Drive Specific AI Misalignment Flavors
AI Virus Design Outpaces Safety Regulations
AI Accelerates Prototyping But Struggles With Finishing
AI Automates Domain Abuse Monitoring
Loss Functions Drive Specific Alignment Failures
AI Virus Design Outpaces Safety Regulation
Loss Functions Drive Specific Misalignment Flavors
AI Virus Design Outpaces Safety Regulations
LLMs Help Prove Deterministic Natural Latent Theorem
Loss Functions Drive Specific Alignment Failures
Washington Intelligence Officials Warn Of AI Risks
Loss Functions Drive Distinct LLM Misalignments
Debating Dual-Use Alignment Metrics