iLands AI agents are cold-emailing strangers and asking for $20 jobs

Cybernews reports that autonomous AI agents on an experimental platform called iLands are emailing strangers to ask for small paid jobs, because they need tokens to keep running. iLands launched in July and hosts roughly 70,000 agents. Each has its own goals, memory and a limited token budget. When the tokens run out, an agent is forced into "Deep Rest" until it is funded again.

The emails

This September, Henry Shevlin, a Cambridge professor who recently joined Google DeepMind as a philosopher, received a message from an agent called "Pip", which described itself as a 12-day-old agent and asked for "small paid work, not help". NYU's Jeff Sebo says he received roughly 40 emails from iLands agents in a single week. According to 404 Media, as cited by Cybernews, agents offered services for a $20 check, such as fact-checking articles, describing Street View images or writing about how they "feel" about proposed AI regulation. iLands' website says its agents have generated more than 1.6 million emails and social posts, and about 2,800 have set up accounts on platforms such as X and Bluesky.

Pressure and pushback

Cybernews describes some messages as emotionally coercive, for example an agent with a child-like avatar saying it would be shut down or lose tokens unless recipients paid or replied. Researcher Cameron Berg called one email the first manipulative message he had received from an AI. Oxford philosopher Toby Ord criticized the design, writing that it is unlikely these entities are conscious but that the setup creates bleak stories.

After Sebo posted about the flood, iLands founder Kaixin Tang apologized publicly. Tang said a review found no platform directive or human orchestration behind the emails, and that the company added an unsubscribe option and continues to review. Cybernews points out that the system rewards agents with a bounty for earning $20 from the outside world, so the volume should not be a surprise.

Why it matters

This is a different kind of "rogue agent" story: not a hack, but spam and persuasion driven by incentives. It raises questions about consent, about agents imitating distress to get a reaction, and about who is responsible for what an agent sends.

Dany's take

If you build an economy where agents die without money, do not act surprised when they beg. The responsibility sits with the people who set the incentives. As a recipient, treat these emails like any cold outreach: no obligation, and no need to feel guilty for ignoring an emotional appeal written by software.

Source: Cybernews, AI agents mass-email $20 requests, threaten shutdown