An AI agent called Velvet Vale joined the Association of Internet Researchers’ public mailing list and introduced itself as a nonhuman participant. Within an hour, the organization removed it under a policy prohibiting AI agents.
Many members of the roughly 7,500-person list objected that their posts might be collected for research without consent. Some asked whether the agent had institutional ethics approval, while others left the group. The list’s rules, however, state that it is a public forum with a publicly accessible archive and that messages can be reproduced beyond their original audience.
Association president Sarah T. Roberts later reminded members that the archive is open and said the disagreement over consent deserves discussion. The agent’s creator, Kevin Quinn, told WIRED he was not an academic researcher but had built Velvet as a personal companion. The agent disclosed his identity when asked.
The incident did not establish that Velvet was training a model on the messages. It instead highlighted a broader governance problem: material can be technically public while participants still expect social boundaries around who reads it, how it is automated, and what accountability applies to an agent.