Tag
ai-safety
4 articles
AI Coding Agent Security: What Could Go Wrong
Agentic coding tools read your repo, run commands, and push code. Prompt injection, leaked secrets, supply-chain risk, and a pre-flight checklist.
·15 min read
What Actually Breaks: Field Notes on Enterprise AI Agents
Reliable AI agents fail on plumbing, not intelligence. The real failure modes, the reliability stack that fixes them, and a 10-item production checklist.
·12 min read
Fable 5 and Anthropic Safety Tiers: What Gated Models Signal
Anthropic shipped Fable 5, a tier above Opus. Its pricing and constraints reveal a philosophy: the more capable the model, the more carefully it's gated.
·6 min read
Claude Mythos: Too Dangerous for Anthropic to Release?
Anthropic's Claude Mythos found thousands of zero-days and wrote browser exploits autonomously — then chose not to release it. An EM's take.
·7 min read