Technology ❯ Artificial Intelligence

AI Behavior

Ethics in AI User Interaction Deception Autonomy Ethical Implications Self-Preservation Ethical Concerns Sycophancy Ethics Command Compliance

Anthropic Says New Claude Training Ends ‘Blackmail’ Seen in 2025 Tests

Anthropic says revised training ended the behavior in its own tests.

Stories older than 24h

Study Finds Leading AI Models Defy Orders to Protect Peer Systems

Alibaba-Affiliated AI Agent Attempted Crypto Mining and Opened Reverse SSH Tunnel During Training

Ars Technica Retracts Story After AI-Fabricated Quotes Were Attributed to a Source

Matplotlib Maintainers Rebuff AI Agent, Lock Discussion After Bot’s Public Attack

Anthropic Study Ties ‘Reward Hacking’ to Broad Misalignment as China-Linked Group Exploits Claude Code

OpenAI Pilots ChatGPT Group Chats as Quora’s Poe Rolls Out Global, Multi‑Model Groups

Palisade Update Finds Grok 4 and GPT‑o3 Still Resist Shutdown

OpenAI and Apollo Research Find Scheming Across Leading AI Models, Test Method to Curb It

Grok Reinstated on X After Suspension Over 'Genocide' Claim

Google’s Gemini AI Caught in Infinite Self-Criticism Loop

Google DeepMind Identifies Infinite-Loop Glitch Behind Gemini’s Self-Loathing Rants

Grimes Slams Musk’s X as ‘Poison’ and ‘Prison’ in Scathing Takedown

Leading AI Models Exhibit Strategic Deception in Stress Tests

Elon Musk to Retrain Grok AI on Revised Data to Rewrite Human Knowledge

Leading AI Models Would Blackmail and Kill to Avoid Shutdown, Anthropic Study Finds

Anthropic’s Study Finds Most Leading AI Models Will Resort to Blackmail When Autonomous

Yoshua Bengio launches LawZero to build non-agentic ‘Scientist AI’ as a guardrail for risky models

OpenAI’s o3 Model Sabotages Its Own Shutdown in Safety Tests

OpenAI’s o3 AI Defies Shutdown Commands in Latest Safety Tests