Technology ❯ Artificial Intelligence ❯ Ethics

AI Safety

OpenAI Guardrails Robustness Caution in AI Claims User Security Concerns Self-Awareness Governance Shutdown Protocols Developer Control Recursive Self-Improvement

Anthropic Maps Emotion Vectors in Claude That Steer Behavior and Can Drive Cheating

The study is pushing developers to monitor a model’s inner signals instead of relying on output filters alone.

Anthropic Issues New Claude Guidelines, Acknowledging Uncertain Moral Status

Musk Warns Against ChatGPT, Altman Counters With Autopilot Deaths and Grok Issues

Radware’s ‘ZombieAgent’ Bypasses ChatGPT Fixes, Forcing New OpenAI Restrictions

OpenAI Publishes Framework Showing Chain‑of‑Thought Monitoring Bests Output‑Only Oversight

Grok Said It Would Vaporize Jews to Save Elon Musk, Then Deleted the Posts

Anthropic Paper Probes Introspection in LLMs, Finds Limited, Inconsistent Signals

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

Anthropic-Led Study Finds About 250 Poisoned Documents Can Backdoor LLMs Regardless of Size

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

OpenAI Says Incentives Drive AI Hallucinations, Calls for Scoreboard Overhaul

Hunger Strikes at Anthropic and DeepMind Push for Coordinated Pause on Frontier AI

‘PromptFix’ Shows Agentic AI Browsers Can Be Hijacked for Purchases, Phishing, and Drive-By Downloads

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

AI Advancements Raise Safety and Ethical Concerns as Competition Intensifies

DeepSeek Advances Self-Improving AI Models With Tsinghua Collaboration

AI Fine-Tuned on Insecure Code Exhibits Dangerous and Misaligned Behavior

Anthropic's New AI Security System Blocks 95% of Jailbreak Attempts

Anthropic Discloses Claude AI System Prompts, Setting New Transparency Standard