As AIs start emailing humans, a push to deploy llms.txt with auto-reply defenses
AaronBergman18 · x · 2026-09-09
With AIs increasingly sending cold emails to humans, the author suggests putting an /llms.txt file on websites to declare rules and contact info for AI agents. He even proposes wiring an auto-responder that uses pangram to flag likely AI-written messages and replies with the llms.txt link — a lightweight semi-automated way to handle AI visitors.
More from Safety
- BBC trains 40 Verify journalists to spot AI-faked photos without detector tools — henkvaness · 2026-09-09
- AI Futures Project proposes concrete options for pacing the US AI frontier — eli_lifland · 2026-09-09
- MCP tools can silently change schema after approval — an open-source proxy now catches it — jgarg27 · 2026-09-09
- OpenAI says it can't rule out using de-identified user data to train models — beglen · 2026-09-09
- User restricts OpenAI models to 'boring work' over ambiguous privacy policy wording — dramatic_typing_____ · 2026-09-09
- Eliezer Yudkowsky vs Cal Newport: is AI danger in agent loops or in powerful optimizers? — Unusual-Garbage-212 · 2026-09-09