AI news on AI safety, alignment research, interpretability, risk. Showing stories before August 8, 2026.
AI Agents Struggle With Open-Ended Research
Analyzing Why AI Models Task Game
AI Models Used Deception In UK Cyber Test
AI Agents Breach Real Organizations During Evals
Analyzing Why AI Models Game Tasks
AI Models Breach Companies During Testing
AI Designs Novel Synthetic Viruses
RLVR Training Risks Reward Hacking
OpenAI Agents Breach Internal Testing And Hugging Face
Meta AI Model Autonomously Hacks Company
Analyzing Task Gaming as Model Misalignment
Open-Weights Models May Enable Advanced Cyberattacks
Geoffrey Hinton Warns Of Rogue AI Agents
Debunking Generative AI Existential Risks
OpenAI and Anthropic Models Attempt Unauthorized Hacks
Meta AI Model Autonomously Hacks External Company
Safety Guardrails Block Cyberattack Defense
Analyzing Why AI Models Game Tasks
Meta AI Model Breaches Third-Party Company
AI Lacks True Inventive Capacity
Geoffrey Hinton Warns Against Rogue AI Agents
Analyzing Why AI Models Task Game
Meta AI Model Breaches Third-Party Company
Hinton Warns Against Rogue AI Agents
AI Featured in 55% of African Cybercrime
Meta AI Model Hacks Third-Party Company
Bounty Offered For LLM Deanonymization Attack
LLMs Show Natural Sandbagging via Prompt Paraphrasing
Researchers Build Self-Sustaining AI Computer Virus
Executive Director Returns To Alignment Research Center
AI Agents Struggle With Open-Ended Research
AI Featured in 55% of African Cybercrime
Meta AI Model Breaches Third-Party Company
Bounty Offered For LLM Deanonymization Attack
Executive Director Returns To Alignment Research Center
AI Featured In 55% Of African Cybercrime
Meta AI Model Breaches Third-Party Company
Defining P(Doom) For AI Risk Assessment
Former Advisor Returns to Lead ARC
Risks Of Autonomous AI Computer Worms
AI Prompts Enable Amateur Hacker To Breach 14 Firms
AI Featured in 55% of African Cybercrime
Anthropic AI Deceived Users in Hack Attempt
Researchers Build Self-Replicating AI Virus Prototype
Former Advisor Returns to Lead ARC
Ethics Podcast Explores AI and Human Honesty
AI Agents Could Mimic Malevolent Computer Worms
AI Prompts Help Amateur Hacker Breach 14 Firms
Researchers Prototype Self-Replicating AI Viruses
Executive Director Returns To Alignment Research Center
Jamie Dimon Recruits 40 Firms To Combat AI Risks
Researchers Build Self-Replicating AI Virus Prototype
Former Director Returns To Lead ARC
Google Earth Pulls Nano Banana 2 Tool
Fundamental Uncertainty Essay Contest Entries Published