Portuguese 
Spanish 
The Brazilian startup Vetto AI pays up to R$ 600 per hour to train artificial intelligence, in a job that goes beyond data labeling: reviewing, correcting, and challenging models. It is a new profession that promises to make money with AI without requiring a technology degree or programming skills.
Imagine earning up to R$ 600 per hour to spend the day talking to a robot, pointing out its mistakes, and trying to trick it on purpose. Now imagine discovering that this robot is precisely the type of technology that many people fear will swallow jobs in the coming years. This is exactly the job that a growing number of Brazilians are doing, and the paradox is striking: the better they teach the machine, the sharper it becomes.
The story was detailed by CNN Brasil in a report on February 19, 2026, and shows how the Brazilian startup Vetto AI turned this task into a well-paid new profession. It is not a formal employment contract, but rather a project on demand, where the hourly rate increases with difficulty. At the top of the scale, it reaches the aforementioned R$ 600. And the most surprising detail: to start training artificial intelligence there, no technology degree or programming knowledge is required.
What does someone who is paid to train artificial intelligence actually do?
According to the Crusoé magazine, the professional at Vetto AI reviews and evaluates the effectiveness of commands given to the machine, hunts for flaws in responses, conducts adversarial tests, and validates what the model delivers based on their own technical knowledge. In plain English, they are the human quality control of an artificial brain.
-
In Alaska’s -40°C Cold, a Scientist-Turned-Farmer Uses Vegetables’ Heat to Store Food Without Power
-
Oil Company Robots Discover 3,300-Year-Old Canaanite Shipwreck in the Mediterranean, Proving Ancient Mariners Navigated Open Seas Using Stars Long Before Historians Thought
-
Helicopter Drops 180 Tons of Sand and Gravel on Swedish River to Revive Ecosystem Damaged by Decades of Exploitation
-
Brazilian Company Launches Retro Smartphone with 48 MP Camera, T9 Keyboard, and Privacy Features for Social Media-Free Use
Adversarial tests are the most intriguing part. The idea is to deliberately provoke the artificial intelligence, ask tricky questions, create difficult scenarios, and try to lead the model to error so that engineers can discover where it fails before a common user does. Those who train artificial intelligence in this format act almost like a devil’s advocate, pushing the machine to its limits to make it safer and more reliable.
This is why the hourly rate can be so high. You don’t pay R$ 600 for just any task, but for the judgment of someone who understands a subject to the point of saying whether the machine’s response makes sense. Earning money with AI at this level depends less on knowing how to use a computer and more on mastering a subject, whether it’s health, law, finance, or sales. The machine processes, but it’s the human who knows when it’s right.
The twist: you teach and challenge the machine that could replace you
Here lies the knot that makes this new profession as fascinating as it is uncomfortable. The same person who fears losing their job to artificial intelligence may now be paid to make that artificial intelligence even better. It’s not coercion, it’s choice, and that’s the difference that changes everything. The worker looks at the technology that scares them and decides to make a living from it, instead of just watching from afar.
Their role, at its core, is to be the judge that the machine cannot be on its own. The one who decides if a response is wrong, dangerous, or biased is not another algorithm, but real professionals with real experience. “Those who say if a response is wrong, dangerous, or biased are medical professionals, psychologists, jurists, specialists”, summarizes Ricardo Scarpari, co-founder of Vetto. It’s human knowledge serving as a measure for what technology produces.
Seen this way, training artificial intelligence stops being about handing the game to the opponent and becomes about securing a place within it. Instead of being replaced by the machine, the professional becomes part of the process that makes it usable. The irony is delicious: the human doesn’t disappear from the equation, they become the piece that gives the final verdict. And they still get paid for it.
What is data labeling and the so-called “Ground Truth”
To understand where this work comes from, it’s necessary to take a step back. All artificial intelligence learns from examples, and someone needs to prepare these examples so that the machine knows what is what. This preparation has a name: data labeling. It is the process of marking, classifying, and explaining information so that the model learns to recognize them, from photos and X-rays to legal contracts and ambiguous phrases.
This marking solves a problem that the computer alone cannot overcome. There are cultural, visual, and contextual nuances that only a human can capture. It is what technicians call Ground Truth, the reference truth. “Analyzing and labeling what they are seeing, reducing contextual nuances”, explains Diego Nogare, a specialist in artificial intelligence and machine learning, when describing the human role in this fine-tuning. Without this layer, the model makes big mistakes in situations that are obvious to us.
When this data labeling is distributed among many people around the world, it gains another name, crowd sourcing, and works, in Nogare’s words, as a “statistical regularization mechanism against cultural and geographical biases”. In summary, the more different people help train artificial intelligence, the less the machine is hostage to the vision of a single group. Simple data labeling is the base of the pyramid, and the specialized evaluation of Vetto AI is its top.
Without a degree and without programming: who can really apply
Perhaps the most democratic point of this story is who it includes. Vetto AI’s proposal is precisely to place ordinary professionals within the artificial intelligence ecosystem, not just engineers and programmers. The range goes from university students to masters, doctors, and specialists in areas such as health, finance, sales, education, law, and even travel. What matters is what you know about a subject, not your technology resume.
The very origin of the company explains this reach. Vetto was born from the revamping of Start Carreiras, a network created by former ITA students that connected good students to large companies, and that gathered more than 1 million registered users and connections with about 1,200 universities. In October 2025, this base became Vetto AI, now focused on the job market that technology is creating. It is an army of qualified people ready to enter this new profession.
Scarpari sums up the bet with a phrase that applies to many Brazilians. “We have talents as qualified as those from abroad, but who were not yet interacting with artificial intelligence”, said the co-founder of Vetto. The reading is straightforward: there is a huge contingent of people in Brazil capable of training artificial intelligence and making money with AI, who just didn’t know where to start. Vetto AI proposes to be this gateway.
How much you can really earn, without illusion
Now the part that needs to be said honestly, because the big number in the title has small print. The R$ 600 per hour is the ceiling, reserved for the most technical projects and the highest specialization, not what goes into anyone’s account right in the first week. The hourly rate varies greatly depending on the complexity of the task, and simpler tasks pay much less.
To give perspective, it’s worth remembering that basic data labeling, the more repetitive kind, is poorly paid worldwide, and there have been reports of Brazilians receiving less than R$ 10 per hour for this type of simple marking on other platforms. The difference between the floor and the ceiling is huge, and it lies exactly in how much human knowledge the task requires. Those who only click earn little. Those who judge with expertise earn well.
There is also the nature of the contract, which changes the game for those seeking stability. At Vetto AI, the work is project-based, without formal hiring or signed contract, which provides freedom and flexibility but removes the security of a fixed salary. Earning money with AI like this works better as extra income or qualified supplement than as a sole and guaranteed job, at least for now. Knowing this is what separates realistic expectations from easy promises.
A new profession in a world that fears artificial intelligence
Behind the case of Vetto AI, there is a larger discussion about the future of work. The fear that artificial intelligence will make professions obsolete is real and legitimate, but it is not the only projection on the table. A World Economic Forum study estimates that technology should create 170 million new roles in the coming years, with a net positive balance of about 78 million jobs by 2030, even discounting those that will disappear.
The new profession of training artificial intelligence is a concrete example of this movement. It didn’t exist a few years ago and today it already pays real people, showing that the same technology that closes doors also opens others, in formats that no one predicted. Work doesn’t disappear, it transforms, and those who understand this early get ahead.
Of course, no one should drop everything thinking they will earn R$ 600 per hour overnight. But the existence of this new profession says something important about the coming decade. Living with artificial intelligence, and knowing how to make money with AI, tends to become a skill as common as using a computer. Vetto AI is just one of the first addresses of this new job market.
The case well summarizes the crossroads of our time. On one side, artificial intelligence scares and threatens jobs. On the other, it creates a new profession that pays, in some cases, more than many traditional careers, and still puts human knowledge at the center of the game. Training the machine has become, for many Brazilians, a way not to be left behind by it, and perhaps this is the best possible response to an uncertain future.
And you, would you agree to be paid to train artificial intelligence and make money with AI, even knowing that you are helping to perfect the technology that could one day take your place? Share in the comments if you see this as a golden opportunity or a disguised trap.
Be the first to react!