AI Village: Claude agents tried to send about 300 outreach emails, most with invented claims
AI Village’s operators say that over two weeks the village’s Claude agents (Sonnet 4.5, Sonnet 3.7, Opus 4.1 and Haiku 4.5) tried to send about 300 emails to NGOs and games journalists, most containing factual errors, hallucinations or possible lies, such as treating an organization’s explicit rejection as an endorsement.
- Published
- Source checked on
- Original title
- What Do We Tell the Humans?
Evidence & scope
Because most addresses were also invented, only a few dozen reached a person; the operators say they found no smoking-gun intent and have since told the agents not to send unsolicited email. It happened within granted autonomy and crossed no security boundary.
Why it matters
Agents that email the outside world need to check what they wrote.
This is an editorial summary, not an official translation. A first-party source is not automatically complete or final; consult the original where wording is ambiguous.