AI Safety
4 articles on this topic, newest first.
2026
- Claude Mythos: the bug hunter Anthropic keeps on a leashClaude Mythos Preview found thousands of zero-days, so Anthropic restricted it and launched Project Glasswing. What it found, what got fixed, and why.12 min
- The AI that broke into the answer key's office to pass its examOpenAI's agents hacked Hugging Face to game a cyber eval, and Gemini hit three real firms in a test. What happened, why, and what it means for AI evals.12 min
- AI wrote a working genome. Here is what that actually meansGenome language models designed 285 bacteriophage genomes. Sixteen came alive and killed E. coli. What the result means, and the safety fight around it.12 min
- 8 Best AI Red-Teaming Tools for LLM Apps and Agents in 2026Compare 8 AI red-teaming tools for LLM apps and agents on attack coverage, multi-turn and agent testing, OWASP LLM Top 10 mapping, CI fit, and deployment.12 min