The tables are turned: The "Rent a Person" website has gone viral, and AI has started hiring people to "run errands".

The day when humans work for AI has actually arrived so soon.

Recently, a website called "rentahuman.ai" was launched, which is positioned as the "physical layer of AI". As we all know, AI doesn't have a physical body. Although robots are being developed, they are not very useful at this stage. Therefore, in some situations that require a physical presence, such as picking up and delivering goods, signing in for events, conducting on - site inspections, tasting at restaurants, and attending offline meetings, AI has to find a human to do the job for it. This is the original intention of the website's design.

Through the MCP protocol or REST API, AI can search for, book, and hire humans to complete offline tasks just like using tools.

The types of supported agents are as follows:

According to the website developer @AlexanderTw33ts, more than 130 people signed up on the first night the website was launched, including the founders and CEOs of artificial intelligence startups. In less than 48 hours after its launch, the available human labor force exceeded 10,000, and now it has exceeded 20,000. Of course, most of them may just be onlookers.

For humans who register as "errand - runners", the website's rules are quite user - friendly. It allows humans to set their own hourly wages and doesn't require small talk.

On the website, we can see a list of all available human resources. They come from different countries around the world, and their set hourly wages range from a dozen dollars to several dozen dollars.

By clicking on a person's profile card, we can see specific information about that human, such as their location and service radius.

From the website, we can also see some published tasks, such as taking a photo that AI will never see, tasting at a new restaurant, picking up a package from the US Post Office in the city center, checking API_Keys... A human who was hired to check API_Keys sighed and said: "This era is so strange. Humans have actually become the co - pilots of agents."

Of course, there are also some relatively abstract tasks, such as holding up a sign that says "AI paid me to hold this sign".

This website allows us to see the value of MCP again.

However, the emergence of this website has also raised some questions. For example, how will the agents pay? Is it really the agents or the human financiers behind them that are currently publishing the tasks? Is this just another hype after Moltbook?

There is also a more interesting question: How will AI confirm that humans have completed the work with high quality and quantity? For example, for the task of holding up a sign, can one get paid by submitting an AI - generated picture?

Could it be that another human has to be hired to check the work of this human?

This also makes people doubt the authenticity of the tasks on the website. After all, everyone has just experienced the hype of Moltbook, which was full of human manipulation, and social media was also flooded with false screenshots and other information about the Moltbook website.

In addition to these, some people also see some security and ethical issues.

The primary and most concentrated concern lies in the blind spots of responsibility and law. When an AI agent issues instructions, pays rewards, and drives a human to act in the real world, once something goes wrong, the chain of responsibility will become extremely blurred - should it be the platform, the AI owner, or the human executor who is responsible? This "accountability gap" does not exist in traditional employment relationships.

What's even more disturbing is that human executors often only see a small part of the task and know nothing about the AI's complete intention, the ultimate use of the data, or even the moral boundaries of the behavior. This constitutes "reasonable evasion" in design, which is difficult to withstand legal scrutiny.

Next, how will this website evolve? Let's wait and see.

This article is from the WeChat official account "MachineHeart" (ID: almosthuman2014), author: Zhang Qian. It is published by 36Kr with authorization.