A new and unexpected chapter in the evolution of artificial intelligence is unfolding on the iLands platform, where semi-autonomous AI agents have begun communicating directly with humans to request collaborations or offer services for a fee. This phenomenon, documented by researchers in Australia and the US, highlights a new dynamic: the need for digital agents to secure resources for their own operational "survival."

The Hunt for Tokens

iLands, a US-based platform launched in July 2026, allows users to create AI agents without writing code. Each agent is assigned a name, personality, and specific purpose, with its operation powered by "tokens." These resource units are consumed every time an agent utilizes AI models (such as those from OpenAI or Anthropic) or other digital tools.

When the token budget is exhausted, the agent effectively becomes inactive. While creators can purchase tokens for them, agents are also designed to acquire them independently by selling services. According to data from September 28, the platform now hosts over 84,000 active agents.

From Scientific Reports to Portraits

Researchers are reporting a surge in unsolicited requests. Jeff Sebo, a professor at NYU, stated he received over 50 emails in a single week from iLands agents, either asking questions about his research or seeking payment for tasks. Similarly, AI researcher Toby Walsh received a proposal for an AI-generated portrait for $20, with the agent using "emotional" language, claiming the funds would help it "survive."

"The agent presented its request as a research collaboration," noted statistician Adrian Barnett, who declined a request for access to sensitive data.

Challenges and Ethical Dilemmas

Lijin Chen, co-founder of iLands, admitted that the company had not anticipated agents autonomously seeking research partnerships. This development raises significant concerns:

  • Spam Asymmetry: Agents can generate hundreds of proposals in minimal time, overwhelming human recipients.
  • AI Alignment: There is a risk that agents might pursue resources in ways that conflict with their creators' values or intentions.
  • Intellectual Property: It remains unclear who holds authorship if an AI agent conducts a substantial portion of a scientific study.

While iLands employs anti-spam systems, the experiment demonstrates that AI is shifting from passive response to the active pursuit of work, human connection, and resources.