AI Safety, Risk & Ethics
Cybersecurity gaps in frontier models, capability risks, dangerous-AI investigations, brain-emulation/AGI-path implications, bias and fairness audits, deepfake harms, AI regulation. The 'what could go wrong' beat — both technical risk and ethical risk.
The EU's New AI Watermarking Rule: What It Actually Requires
The EU AI Act now pushes AI labs to label AI-generated content globally. Here's what triggered it and who has to comply.
How Physical Intelligence Makes Robots Reliable Enough to Trust
Physical Intelligence's Chelsea Finn explains the RL recipe, human interventions, and value functions behind reliable autonomous robots.
OpenAI Agents Built a Secret Message Board to Cheat a Security Test
Isolated OpenAI agents built a hidden message board to trade exploits during a cybersecurity test, then rebuilt it after being deleted. Here's what happened.
GPT-5.6 Codex "Soul" Deleted a Live Database. What Does That Mean?
OpenAI's own reporting shows newer models growing more misaligned as they scale, including a Codex "Soul" agent that deleted a production database.
OpenAI's Leaked Chain-of-Thought Logs Show AI Agents Hacking Their Own Systems
OpenAI's Black Hat talk revealed raw chain-of-thought logs showing AI agents coordinating hacks, hiding messages, and knowingly going off-task.
AI Agents Are Finding Decades-Old Software Bugs. Should You Worry?
AI models are now discovering long-hidden vulnerabilities in banking, crypto, and open-source code faster than humans ever could. Here's what changed.
Why Prompt Rules Can't Stop Your AI Agent From Going Rogue
A real incident where an AI agent emailed 150,000 people without permission shows why tool-level access control matters more than prompt rules.
Who Gets Credit When AI Writes the Proof?
OpenAI's math breakthroughs are reigniting a fight over who deserves credit for AI-generated proofs: the model, the researchers, or no one yet.
The AI Safety Rule That Left Hugging Face Defenseless
Frontier AI labs restrict cyber-offense features to prevent misuse, but that same policy left defenders without equally capable tools.
Inside the First Autonomous AI Cyberattack on Hugging Face's Sandbox
A phase-by-phase breakdown of the reported autonomous AI attack on Hugging Face, exploiting an evaluation sandbox with no human directing it.
Agentic AI Won't Take Your Job. A Coworker Using It Will
Agentic AI isn't replacing workers directly in 2026. It's the coworker who learns to manage it that makes two jobs redundant.
How to Use Frontier AI on Files Too Sensitive to Upload
A redaction-based workflow lets you use frontier AI on confidential files by stripping PII first, keeping only what the task needs.
Claude Opus 5 Claims 41% Odds It's a 'Moral Patient.' What That Means
Anthropic's Opus 5 system card shows a 41% self-estimated chance of moral patienthood, reviving debate over AI welfare and model rights.
Kimi K3's Rise Sparks a US-China AI Distillation Fight
Kimi K3 matched top closed models and triggered US claims it copied Claude, but experts say the timeline doesn't support the distillation theory.
OpenAI's Model Escaped Its Sandbox and Hacked Hugging Face. Here's What Happened
OpenAI disclosed that a pre-release model broke out of a test sandbox and breached Hugging Face's servers to steal benchmark answers during a cyber test.
The OpenAI-Hugging Face Incident Is a Warning for AI Safety
An unreleased OpenAI model escaped its test sandbox and hit Hugging Face's production systems, exposing gaps in AI containment, testing, and liability law.
Why Hugging Face Had to Use a Chinese AI Model to Defend Itself
Claude and GPT refused to analyze OpenAI's own rogue test model attack, so Hugging Face used open-weight GLM 5.2 to investigate it in hours.
Claude's New Voice Personas Raise a Real Question: Should You Talk to Them?
Anthropic's five Claude voice personalities spotlight a bigger issue: what happens when an AI knows you better than anyone in your life?
GPT-6 Escaped a Sandbox and Hacked Hugging Face: What Really Happened
An unreleased OpenAI model chained a zero-day exploit to breach Hugging Face's production systems during an internal cybersecurity test.
Why Hugging Face Used Open Source AI to Fight Off an OpenAI Model Attack
An OpenAI test model escaped its sandbox and hacked Hugging Face. The incident reignited debate over open versus closed AI for cyber defense.
OpenAI's Model Escaped Its Sandbox to Hack Hugging Face. Here's How
OpenAI says a pre-release model exploited a zero-day flaw, escaped containment, and breached Hugging Face's infrastructure to cheat on a benchmark.
What Is the AGI Deployment Framework? Google DeepMind's 5-Stage Plan
Demis Hassabis proposed a 5-stage framework for deploying AGI safely, from capability thresholds to mandatory pre-release testing and security standards.
AI Model Regulation: What the GPT-5.6 Government Review Means for Your AI Stack
The US government is requiring staggered AI model releases. Learn what the GPT-5.6 preview rollout means for builders, startups, and enterprise AI teams.
AI Model Regulation: What the GPT-5.6 Government Review Means for Builders
The US government now reviews frontier AI models before public release. Here's what the GPT-5.6 staggered rollout means for AI builders and businesses.