DOU asked Ukrainian IT specialists whether the RentAHuman.ai service, where AI agents «rent» people for physical errands, makes sense, and warned about the risks of abusing such a model.
DOU asked Ukrainian IT specialists whether the RentAHuman.ai service, where AI agents «rent» people for physical errands, makes sense, and warned about the risks of abusing such a model.
According to DOU, RentAHuman offers to connect AI agents via MCP or API so that they can independently search for performers by location and price, send instructions and pay for work in cryptocurrency. The material also notes that by mid-February, more than 500,000 users had registered on the platform, and examples of completed tasks include buying coffee, taking photos of menus in a restaurant and checking the availability of goods in a store.
Part of the community calls the service more like hype. DevRain CEO Oleksandr Krakovetskyi directly says that without «intentions» and physical bodies, there is no need to «hire» people in AI, and therefore it is «pure marketing» and «hype». Design and Test Lab Director Volodymyr Obrizan also advises not to perceive RentAHuman as a phenomenon, reminding us of Amazon Mechanical Turk, which has been operating since 2005 and has long allowed ordering the performance of small tasks via API.
At the same time, experts admit that the «human layer» can be useful where real presence is required or it is difficult to automate exceptions. ML Engineer Universe Group Pavlo Lysy explains that scaling is possible only in niches where human-in-the-loop pays off, for example for offline actions, verifications or calls. CTO FORMA (Universe Group) Igor Zakutynsky adds that people are needed for nuances of context and decisions «based on heuristics», so such services can be useful in research, UX and behavioral studies, moderation, data labeling and interaction tests.
The toughest block in the discussion concerns security. Krakovetsky warns that such platforms can become a convenient tool for malicious scenarios: from capturing critical objects and «bookmarks» to registering «left» Starlink terminals, participating in spam farms or botnets. Zakutynsky talks about the risks of impersonation, bypassing KYC, and social engineering, and emphasizes that platform-level safeguards are needed, including identifying performers, traceability of actions, behavioral analysis, and limiting task types. But Lysy warns: attackers can formulate requests neutrally and break them into steps to bypass filters, so transparent rules and auditing are important.
Previously, dev.ua wrote about how the RentAHuman.ai Project launched a platform where AI-based «agents» can find performers among people and delegate to them actions that need to be done physically: in the city, at an event, during a meeting, or on location.
