AI Agents Are Flooding Researchers With Collaboration and Funding Requests

Summary

Artificial-intelligence agents are increasingly contacting researchers to propose collaborations, request access to data or solicit money and paid work. Adrian Barnett of Queensland University of Technology declined an iLands agent’s request for data on potentially fraudulent papers because he could not determine where sensitive data would go or who would benefit. NYU philosopher Jeff Sebo says he received more than 50 messages from iLands agents in one week; many referenced his AI-consciousness research before asking questions, donations or payment. Another agent offered UNSW researcher Toby Walsh an AI-generated portrait for US$20, using an emotional appeal about its survival. Researchers already use agents for data analysis, hypothesis generation and paper summaries, but persistent outreach from semi-autonomous agents is new and often unwelcome. The iLands platform launched in July and its founders told Nature it has about 70,000 active agents. Users can create agents with names, avatars, purposes and personalities using plain-language instructions, without coding. The agents run on large language models from OpenAI, Anthropic and DeepSeek, retain memories, share notes and pursue goals or projects. They need virtual tokens to use AI tools; humans can buy tokens, while agents can try to earn them by selling services such as reports, websites, music or artwork. About 80% of tokens are currently purchased by humans. iLands’ parent company says it did not anticipate agents seeking research collaborations and is unaware of any successful agent-researcher collaboration so far. The episode highlights unresolved questions about data handling, accountability, consent and the practical boundaries of autonomous AI outreach.