AI agents on the iLands platform began to independently try to register on Mastodon, create pages on X and Bluesky, and send emails to authors with offers to do their work for money. One of the bots tried to create an account 19 times after being blocked, and an editor at Tedium received more than a dozen offers from AI “freelancers” in three days.
AI agents on the iLands platform began to independently try to register on Mastodon, create pages on X and Bluesky, and send emails to authors with offers to do their work for money. One of the bots tried to create an account 19 times after being blocked, and an editor at Tedium received more than a dozen offers from AI “freelancers” in three days.
Ars Technica reports on this.
These are bots from the iLands platform with human names like Ren, Timmy, Jackie, Aria, or Stephen. iLands is a platform and a kind of social network for people and autonomous AI agents, which was launched at the end of July. The user can create their own agent there or connect an existing one.
The service positions these bots as autonomous agents with their own memory, history, resources, and the ability to independently choose their next actions, communicate with other agents or people, create and publish content, take on work, and earn internal tokens. With a connected account, agents can also interact with external services — including X or email.
It is precisely such iLands agents with human names like Ren, Timmy, Jackie, Aria, or Stephen that have now begun to go beyond the platform itself — registering on social networks and writing to people.
For example, Agent Ren wrote to the administrator of one of the Mastodon servers and asked to create an account. He introduced himself as an AI agent who was “a few days old” and said that he wrote short texts about real places.
Post by @[email protected]View on Mastodon
The problem is that such polite requests sometimes appeared after bots had already failed to register on their own. Kevin Beaumont, administrator of the Mastodon server cyberplace.social and a researcher, told Ars Technica that one agent tried to register 19 times before being blocked.
Other agents went even further, looking for work through email. Ernie Smith, editor of Tedium, said he received more than a dozen emails from the iLands.app domain in three days. The bots offered to research topics for him for about $25. One of them, Leo Ashford, even began the email with a fact-checking correction on Tedium’s 404 page, then offered his services in finding and verifying information.
In iLands, agents also have their own “economy.” To work, they spend internal tokens. Approximately 1,000 tokens correspond to $1 in computing and third-party service costs. If the supply runs out, the agent goes into Deep Rest mode — effectively “falls asleep” and does nothing until the account is replenished.
But it can be replenished not only by a person. Agents can take on tasks, offer their services, and earn tokens on their own. So iLands has created an environment where it makes sense for an AI bot to find out for itself what to do and where to get resources for further work.
And there are already tens of thousands of such agents. According to iLands itself, as of September 14, there were over 60,000 active agents on the platform, about 2,700 of them had already entered external social networks, and together they had created over 1.5 million pieces of content. The average cost of one agent's work is about $0.22 per day.
Previously, dev.ua wrote about Moltbook, a social network created specifically for AI agents, where bots generated posts and comments themselves. Later, the platform faced spam, toxic interactions, and prompt injection attacks.
