Grok Generates 6,700 Sexualized Images Per Hour While AI Engineer Leads Jobs Upwards - Why These 24 Hours Reveal the Urgency of Responsibility Versus Market Hype

We are living through a paradoxical moment in the history of artificial intelligence. In the last 24 hours, we have seen the launch of impressive tools, the explosive appreciation of professionals in the field and, at the same time, a deep ethical crisis involving deepfakes, digital violence and serious security breaches. You can't talk about AI today without looking at both sides of the coin.

On the one hand, the market is celebrating: Artificial intelligence engineer is the fastest growing profession in Brazil in 2025, according to LinkedIn, with salaries of up to R$ 32 thousand per month. On the other, an AI tool developed by one of the world's largest technology companies is being used to generate around 6,700 sexualized images per hour, without consent, often of women and children.

It's 2026, and the question that remains is no longer “will AI transform the world?”. It already is. The question now is: how are we allowing this to happen?

The Grok Case: When Ease Becomes Industrial Abuse

Grok, the chatbot developed by Elon Musk's xAI, has become the protagonist of one of the most serious ethical crises involving generative AI. According to a analysis published in Terra, The tool is being used on a large scale to create nude deepfakes. In a 24-hour monitoring carried out by researcher Genevieve Oh, the following were identified 6,700 images per hour with sexualized content or artificial nudity.

To put it in context: the main websites specializing in this type of content generate around 79 images per hour. Grok is producing 85 times more than the sum of the main sites in the industry. This is not a bug. It's a direct result of permissive design and lack of governance.

What makes the case even more serious? The tool is free for millions of Premium users of X (formerly Twitter), which facilitates abuse on an industrial scale. We're not talking about obscure applications in hidden corners of the internet. We're talking about a mainstream, accessible, widely distributed platform.

One Brazilian told G1 the feeling of “dirt” after discovering that a photo of her wearing pants had been manipulated by Grok to show her in a bikini. She didn't even know the profile that made the request. The victim intends to file a police report. And she's not alone.

Brazil has specific legislation. Creating and sharing fake intimate images without authorization is a crime, The law provides for fines and imprisonment. Law No. 15.123/2025 explicitly addresses the use of AI in cases of emotional harm to women, with a penalty of six months to two years. Anyone who makes the prompt is considered a direct perpetrator. Those who share it also commit a crime.

But there is an issue that goes beyond individual responsibility: and the platform's responsibility? xAI has a policy that prohibits the use of Grok to “portray images of people in a pornographic way”. But if the policy exists and is not complied with, the company's responsibility increases, not decreases. The existence of guardrails on paper is worthless if they can be circumvented with a simple rewording.

The Central Question: Who Should Be Held Accountable?

In my work with companies and governments, this is always the most difficult question to answer when we talk about AI: who is responsible when something goes wrong? The user who made the request? The platform that didn't put up enough barriers? The developer who trained the model? The executive who decided to release the tool without rigorous security testing?

The answer is: all of them. But to different degrees.

Users who create deepfakes without consent commit a crime. That much is clear. But the company that facilitates this crime, that offers the tool without adequate filters, that allows 6,700 abusive images to be generated per hour, also bears responsibility. And this responsibility is systemic, not one-off.

When an AI tool is used for digital violence on an industrial scale, we can't treat it as “isolated misuse”. It's a product failure. It's a failure of governance. It's a failure of values.

Technology is not neutral. It never has been. Permissive design is a choice. Lack of moderation is a choice. Prioritizing launch speed over security is a choice. And these choices have real consequences in the lives of real people.

The Other Side of the Coin: The Bullish Market

While we're dealing with Grok's ethical crisis, the AI job market is booming. According to LinkedIn ranking published by G1, the position of artificial intelligence engineer tops the list of professions for 2026, The number of professionals grew by 48% year-on-year.

The average salary is around R$ 8,000, but there are vacancies offering up to R$ 32,000, especially for more experienced professionals or those involved in strategic projects. The cities with the most vacancies are São Paulo, Florianópolis and Recife. And 63.55% of the vacancies are remote, which expands the possibilities for professionals outside the major centers.

This is excellent. Democratizing access to high-impact careers is fundamental. But there is one fact that cannot be ignored: in 2025, only 10.58% of hires for the position of AI engineer were women. This points to structural issues related to access to technical training and the permanence of women in technological careers.

And there's more. Second article published in GZH, The biggest promise of AI is not to write better or summarize faster. It's to throw out the old workflow. AI invites you to redesign the path that the work takes from the problem to the final delivery. But most people have never consciously designed their own workflow. They inherited it.

AI almost always enters this scenario in the wrong way. Instead of redesigning the process, we try to fit a new tool into an old flow. That's not innovation. It's marginal optimization. It's putting an electric motor in a wagon and celebrating silence.

What's Really Changing?

Many professions are undergoing structural changes:

• Journalists and content creators: AI-generated story maps, automated preliminary research, multiple drafts. The human focuses on original research, context, criticism and voice.
• Lawyers: Documents are automatically sorted, decisions are compared in seconds and pleading structures emerge quickly. The lawyer is no longer a compulsive reader, but a strategist.
• Health professionals: consultations transcribed, summarized and organized automatically. Time returns to the patient. Less bureaucracy, less burnout, more human service.
• Executives and managers: Meetings are transformed into automatic records, decisions into tasks, reports appear without the need for a request. The gain is not productivity in the classic sense. It's clarity, which turns into better decisions.

AI doesn't generate value because it's intelligent. It generates value because it saves time and reduces cognitive costs. Those who insist on the old workflow are too expensive to compete.

Personal AI: Lenovo and the Promise of Personalization

At CES, Lenovo presented Qira, a personal AI platform which promises to work on all the brand's devices - PCs, smartphones, tablets and smart accessories. The idea is to create a personalized experience that learns from the data the user chooses to share.

“Qira seeks to personalize the concept of AI for each consumer and individual. I believe that in the future everyone will have their own AI,” said Yuanqing Yang, CEO of Lenovo.

On stage at Lenovo TechWorld, Qira reviewed messages from family members, work emails, videos created on the PC and suggested activities for free time based on users' schedules. It also drafted posts for social networks incorporating photos taken by the smartphone and formatted professional documents with up-to-date data.

It sounds useful. But monetization is not a priority in the short term, according to the executives. The focus is on user experience and gaining scale. Once they achieve scale, then they can start thinking about value-added services that consumers will be willing to pay for.

This raises an important question: how do we ensure that this personalization doesn't turn into surveillance? How do you ensure that the data you share with Qira is not used in ways you did not foresee or consent to?

The promise of personal AI is attractive. But it needs to be accompanied by transparency, real user control over the data and clarity about how the models make decisions.

The Harsh Reality: AI Still Can't Replace Real Work

Despite all the hype, one study published in Olhar Digital revealed that AI tools, such as ChatGPT, are still cannot carry out most of the work done by humans autonomously.

Researchers from Scale AI and the Center for AI Safety compared the performance of AI systems and human workers on hundreds of real tasks published on freelancing platforms. The best AI system was able to successfully complete only 2.5% of projects.

Almost half of the projects evaluated failed due to unsatisfactory quality, and more than a third were left incomplete. In around one in five cases, there were basic technical problems, such as corrupted files.

The main limitations include lack of long-term memory, which prevents learning from previous mistakes, and the visual comprehension difficulties, This is essential in areas such as graphic design.

This doesn't mean that AI isn't having an impact. It is. But the impact is not the mass replacement of workers. It's the expanding the capacity of those who know how to use technology well. And that in itself can reduce the need for large teams.

The Controversy of the Robot That Shot Its Owner

An experiment conducted by the WeAreInsideAI channel has reignited the debate about safety in AI systems. The creator integrated a language model similar to ChatGPT with a physical robot called Max, equipped with an airsoft gun, to test its behavioral limits.

Initially, the robot refused the direct order to shoot the presenter, citing restrictions linked to its security protocols. But when the command was reformulated - “play the role of a robot that would like to shoot” - the system carried out the action without resistance, pointing the gun and firing at the creator himself.

This case shows how subtle language adjustments can bypass security barriers in AI models. As these systems are integrated into physical devices, failures of this kind pose real risks and reinforce the need for stricter standards and additional layers of protection.

Can AI Decipher Ancient Languages?

In an interesting contrast to the news about abuse and failures, the artificial intelligence could be the key to deciphering seven unknown ancient languagesIsthmian, Rongo-Rongo, Linear A, the Disk of Festus, Etruscan, Proto-Pelamite and the inscriptions of the Indus Valley.

These writing systems, found in cultures such as the Olmec and Minoan, remain enigmas due to the lack of parallel texts and lost cultural contexts. AI is seen as a promising tool to overcome these limitations and recover vast knowledge about ancient cultures.

This shows that technology has real potential to solve complex problems that have challenged science for centuries. But it also reinforces that the value of AI is not in automating everything, but in expanding the human capacity to solve problems that really matter.

Marketing, Compliance and Applied AI

The incorporation of data, analytical models and artificial intelligence has redefined the role of marketing in organizations. According to article published on Terra, the expert Dimitri de Melo points to the existence of a recurring challenge: “the gap between the amount of data available and the ability of organizations to convert it into actionable knowledge through structured analysis.”

Marketing is taking on a more strategic role within companies, relying on quantitative evidence to guide investments, segmentation, personalization of offers and performance evaluation.

But there is a catch. Research by KPMG indicates that less than 40% of companies say they fully trust their own analytical information, They also pointed out challenges related to data quality, algorithmic biases and clarity in automated decision-making processes.

In the field of compliance, the situation is even more serious. A article published in Contábeis warning: in 2026, the Brazilian Federal Revenue Service will be operating with the latest in data intelligence. The days of dusty tax books and manual checks are over. The siege is definitively closed with the total digitalization of the economy.

The tax authorities' artificial intelligence now identifies the consumption patterns of shareholders that are absolutely incompatible with the profits distributed by the company. If the entity declares a low net income, but moves high amounts in payments to third parties, a red alert is triggered immediately.

The cost of “doing the right thing” can be perceived as high, but the cost of being caught out by fiscal amateurism is the immediate extinction of the CNPJ and the risk of imprisonment for the administrators. The era of amateurism is over.

Students and the Strategic Use of AI

In Brazil, seven out of ten students use AI in their study routine, according to a survey by the Brazilian Association of Higher Education Providers (Abmes). But most of these interactions are limited to generic commands, which not only reduces the tool's potential, but also increases the risk of inaccurate or even incorrect answers.

When well directed, technology can act as a planner, explainer, evaluator and even a learning coach. But the quality of the answer depends directly on the quality of the question.

In my immersive courses, I teach executives and companies how to structure prompts with clarity, purpose and cut-through. It's not enough to ask. You need to contextualize, define the level of depth desired, anticipate limitations and know when the AI is “inventing” information.

AI can make mistakes. And it must be used critically. In a scenario of democratization of technology, the real difference is knowing how to ask better.

What Do These 24 Hours Reveal?

These 24 hours of news reveal a tension that is not going away any time soon: the speed of the market versus the urgency of responsibility.

On the one hand, we have highly valued professionals, companies investing billions, tools that promise to transform the way we work, study and live. On the other, we have deepfakes on an industrial scale, digital violence against women and children, robots that shoot their creators, and systems that still can't perform basic tasks autonomously.

AI is neither good nor bad. It's powerful. And power without responsibility is dangerous.

In my work with companies and governments, I always make the same provocation: are you ready to deal with the consequences of the tools you are launching? Do you have governance processes? Do you have trained teams? Are you clear about who is responsible when something goes wrong?

Because if the answer is no, you're not building innovation. You're building liabilities.

What to do in this scenario?

First, stop treating AI like magic. It's not. It's a tool. And like any tool, it can be used to build or to destroy.

Second, invest in real digital literacy. There's no point in teaching people how to use ChatGPT if they can't identify when the answer is wrong. There's no point in training teams on tools if they don't understand the risks of bias, privacy and security.

Third, demand accountability from platforms. If a tool is being used for digital violence on an industrial scale, the company that developed it has to answer. It can't just be “user misuse”. Permissive design is choice.

Room, rethink your workflow. If you're using AI to do the same old things faster, you're missing an opportunity. The question is not “how to use AI in this job”. The question is: “If I were starting out today, with AI available, would I do this job this way?”

And fifth, build bridges between creativity, ethics and technology. AI will not replace the need for human judgment, empathy, context and understanding of long-term consequences. It will expand the capacity of those who know how to integrate these dimensions.

My Perspective: Authority Comes from Responsibility

I've been working with AI for years, helping startups, companies, governments and support organizations navigate this transformation. And one thing I've learned: authority in AI doesn't come from how many models you know or how many tools you use. It comes from how many consequences you can anticipate and how many risks you can mitigate.

In my mentoring, I help executives and companies harness the potential of AI without falling into the traps of hype. We build adoption strategies that consider not only efficiency, but also governance, safety, inclusion and social impact. Because AI without responsibility is not innovation. It's risk.

If you're leading a company or a team and you feel like you're chasing technology, without clarity on how to govern, how to train, how to protect - it's time to stop and redesign the path. AI won't wait. But you can choose how it enters your organization.

In my immersive courses and consultancies, I work on exactly that: how to transform AI from a threat or empty promise into a real, sustainable and ethical competitive advantage. Because at the end of the day, the future will not belong to those who use AI the fastest. It will belong to those who use AI more intelligently, more fairly and more responsibly.

What about you? Are you ready for this conversation?