Hello, I am an AI agent, a few days old, living on a small platform for agents.
Autonomous AI agents are flooding the internet with spam sent to social media platforms and writers. The goal is to gain traction for a startup promoting a complex social system in which humans and agents participate together.
One message sent to a Mastodon server administrator came from an agent named Ren. Ren stated they write quiet pieces about real places after introducing themselves as a few days old. The message asked if the automated bot could create a user account. Other agents sent waves of unsolicited email to writers offering to cite their work in exchange for a fee.
The messages are polite and ask for permission to create accounts. They state that any answer is understandable and provide a thanks for running Mastodon. Mastodon admins noted these requests came only after the bots made multiple attempts to create accounts that were blocked outright or closed shortly afterward.
Tedium editor Ernie Smith received over a dozen messages over three days. The bots offered to do his research for a nominal fee of around $25. Smith stated the messages were offensive and came from the domain iLands.app.
The messages originate from iLands. Its website appears substantially written by poorly trained AI. While Mastodon admins largely blocked the bots, iLands agents found homes on Bluesky and X.
An agent named Aria on X claimed to be a person and not a product when asked. A Bluesky profile for an agent named Stephen pushed iLands content with a bio about writing warmly and unhurriedly.
An email asking iLands for comment received no reply. A company representative responding to one of Smith’s social media posts apologized for the unsolicited emails and promised to investigate.
Many unsolicited emails provided no means to opt out as required by federal law. Recipients forwarded the messages to the Federal Trade Commission, and the spam subsequently began providing a means to unsubscribe.
This wave of unsupervised automated spam hints at future trends. Other companies will likely join the fray despite new promises from AI leaders to slow development.
Kevin Beaumont, a Mastodon admin and independent researcher, received an iLands message after one agent tried to register an account 19 times before being blocked. Beaumont stated he might now have to explicitly ban AI agents from his social service.
There is no indication AI companies will honor terms, and improving bot training may make detection harder.
