Anthropic Cuts Fable 5 Biology Blocks by 85%
Anthropic slashed Fable 5 biology-related fallbacks by 85% with smarter safety classifiers. Here’s what changed, how it works, and what it means for healthcare.
AI Safety
Read More
OpenAI and APA Team Up on Youth Mental Health AI Safety
OpenAI and the American Psychological Association are building evidence-based AI safeguards for young users. Here’s what the partnership actually involves.
OpenAI’s Cybersecurity Evaluation Incident: What Went Wrong
OpenAI disclosed a third-party cybersecurity evaluation incident involving its models. Here’s what happened, what it means, and what safeguards are coming.
OpenAI Shut Down a Cambodia Scam Ring Using ChatGPT
OpenAI disrupted a Cambodia-based criminal operation using ChatGPT for romance scams, fake investments, and impersonation. Here’s what actually happened.
OpenAI’s EU Compliance Playbook: What It Actually Means
OpenAI outlines its safety, transparency, and provenance practices for Europe. Here’s what the EU AI Act actually demands — and how OpenAI plans to meet it.
OpenAI and Hugging Face Hit by Security Incident During Model Eval
OpenAI and Hugging Face just disclosed a serious security incident during AI model evaluation. Here’s what happened, what it reveals, and what defenders must know.
OpenAI’s Long-Horizon AI Safety Lessons Are a Wake-Up Call
OpenAI reveals new safety risks from long-running AI agents — real failures, surprising behaviors, and the safeguards they’re building to keep agentic AI in check.
OpenAI Wants Teens on ChatGPT — Here’s How It Plans to Keep Them Safe
OpenAI is rolling out age-appropriate protections, parental controls, and expert-backed safeguards for teens on ChatGPT. Here’s what’s actually changing.
GPT-Red: OpenAI’s Self-Improving AI Safety System
OpenAI’s GPT-Red uses self-play to automate red teaming for AI safety. Here’s what it does, why it matters, and what it means for the future of alignment.
OpenAI’s Reverse Federalism Bet on AI Safety
OpenAI wants state laws to build the foundation for federal AI safety rules. Here’s what their ‘reverse federalism’ approach actually means and why it matters.
Anthropic Is Asking the Public Hard Questions About AI
Anthropic’s new ‘hard questions’ initiative surveyed 133,000 people across 159 countries. Here’s what they found — and what they’re promising to do about it.
OpenAI’s GPT-5.5 Bio Bug Bounty: What It Actually Is
OpenAI is paying researchers to break GPT-5.5’s biosecurity guardrails. Here’s what the Bio Bug Bounty program covers, who can join, and why it matters.