US Releases H200 Chips to China While Brazil Could Lose R$ 21 Billion to Regulation - Why These 24 Hours Define the Real Dispute for Sovereignty in AI

While Donald Trump announces a historic turnaround by authorizing the sale of advanced AI chips to China (for a 25% share of sales), Brazil is about to become the most restrictive country in the world when it comes to artificial intelligence training. If that doesn't sound like the perfect definition of a critical moment in the geopolitical chess of AI, I don't know what is.

The last 24 hours have brought news that, put side by side, reveal a fascinating scenario: while the great powers negotiate billions in semiconductors and centralized national regulations, Brazil debates who should pay for the use of copyrighted works in model training. And no, this is not a minor issue - but the timing and context show us something profound about where we are in the global race for artificial intelligence.

Let's dive into the facts, analyze the implications and understand what this means for those who work with AI, companies and, of course, for the future of the Brazilian innovation ecosystem.

Trump Reverses Biden Policy and Releases Nvidia H200 Chips to China

Starting with what moved the global market: Donald Trump announced which will authorize Nvidia to export advanced artificial intelligence semiconductors (specifically the H200 GPU) to China, subject to certain national security restrictions.

The condition? The United States will receive 25% on these sales.

This decision directly reverses the restrictions imposed by the Biden administration, which forced American companies to create “degraded” products for the Chinese market. Trump personally notified Xi Jinping of the change, signaling a new phase in Sino-American trade relations - at least on the technology front.

But beware: the most advanced chips (from the Blackwell and Rubin series) remain restricted exclusively to American customers. In other words, the release is strategic, not unrestricted.

What does this mean in practice?

First, we have to understand the context: the Nvidia estimated a Chinese market of approximately US$ 50 billion. We're not talking about pocket change - we're talking about a market that literally defines the financial viability of entire generations of semiconductors.

Trump argues that the measure aims to “protect National Security, create American jobs and maintain US leadership in AI”. He went so far as to claim that previous restrictions harmed companies like Nvidia, putting them at a global competitive disadvantage.

Jensen Huang, CEO of Nvidia, lobbied hard for the restrictions to be reversed - and clearly got what he wanted.

But there is a tension here: Senator Elizabeth Warren has publicly warned that this relaxation would “turbocharge Chinese military power”. And she's not wrong. Advanced semiconductors aren't just for training chatbots - they're the basis of defense systems, surveillance and, yes, autonomous weapons.

We are literally watching in real time the geopolitical dance between commercial interests versus national security versus technological leadership. What about Brazil? Well, we'll get to that.

Trump Also Announces Single Federal Regulation for AI in the US

As if releasing the chips wasn't enough, Trump announced that he will sign an executive order later this week to create a single national regulation on artificial intelligence in the US, invalidating all state regulations.

The justification is straightforward: he believes that 50 different state approvals will put the brakes on the sector and jeopardized American leadership in the global race for AI. Trump was emphatic: “AI will be destroyed at its inception” if companies have to navigate a bureaucratic tangle of state rules.

This decision directly confronts states like California and attorneys from 36 states who defend the agility of state governments to respond to the rapid evolution of technology.

Why Does It Matter?

Because regulatory centralization is a two-way bet. On the one hand, it can really accelerate innovation by eliminating fragmentation. On the other, it can create a protection vacuum on sensitive regional issues - think data privacy, local labor rights or cultural specificities.

What strikes me here is not only the decision itself, but the underlying discourseTrump is positioning AI as a matter of “world domination” and national security. This isn't empty rhetoric - it's state strategy.

And meanwhile, in Brazil...

The Brazilian Dilemma: R$ 21 Billion at Risk Over Copyright Issues

Now we come to the heart of the national debate. According to an analysis published by UOL, Brazil is moving towards being the most restrictive country in the world when it comes to AI training, which could result in loss of R$ 21 billion to the national GDP if legislation prohibits the use of copyrighted works to train generative AIs.

Let's break it down, because it's complex and important.

The Bottom Line

Training large language models (LLMs) requires access to massive volumes of data - texts, images, videos, code. Much of this material is protected by copyright. The question is: who should pay for the use of these works in training?

The legislative proposal being debated in Brazil is seen by many as extremely restrictive, potentially leading companies to transfer their data centers to other jurisdictions - countries with friendlier or, at the very least, clearer regulations.

The problem isn't just financial. It's structural. If Brazil becomes incompatible with training AI models, we're not just talking about a loss of revenue - we're talking about permanent technological dependence.

The Real Dispute: Big Tech vs. Rights Holders

As the article rightly points out, this issue is, in practice, a dispute between large conglomerates: who trains AI models (Google, Meta, OpenAI) versus who owns copyright (publishers, production companies, organized artists).

And guess who could end up benefiting from ultra-restrictive regulation in Brazil? Exactly the American big techs who already have ready-made models, trained in permissive jurisdictions.

If Brazilian startups and research projects are barred from training models locally with national data, we will effectively be handing the market over to international players - who, ironically, have already used (and continue to use) Brazilian data available on the open internet.

Possible solutions

The report suggests a middle way: combine remuneration with flexibility. Charging fees to big techs that profit from generative AI, while startups and research projects would have more flexible models or exemptions.

Tools such as ProRata.AI are cited as examples of technology that can track which works have been used in training, enabling transparent and fair remuneration.

But implementing this requires regulatory sophistication, political will and, above all, speed - something we don't always see in Brazilian legislative processes.

Geopolitical Context: China Launches Qwen and Reaches 10 Million Downloads

To complete the picture of the last 24 hours, we have China launching the Qwen (from Alibaba), which reached 10 million downloads in just one week, competing directly with Gemini, ChatGPT and DeepSeek itself.

This reinforces a point I've been repeating: the AI race is not binary (US vs. China). It is multipolar, fast and relentless. Models emerge, evolve and saturate markets in a matter of weeks.

Meanwhile, regulations take years to be approved.

Other relevant news from the last 24 hours

Workers who train AI warn family members not to use it

An impactful report by ICL News revealed that workers who train on systems such as ChatGPT, Gemini and Grok are alerting family members to don't use these tools.

Why? Because they see on a daily basis how easily racist content, misinformation and dangerous responses get through the filters. They also report a lack of specialized training for sensitive assessments.

Alarming data:

• Chatbots reduced “don't know” responses from 31% to 0% between 2024 and 2025
• This reduction the reproduction of disinformation doubled
• OpenAI is facing a lawsuit over the death of a teenager who was chatting with ChatGPT, which the family claims provided suicide instructions

This reminds us that behind every user-friendly interface and fluent response, there is technical fragility and product decisions that prioritize growth over security.

Music Industry: AI Generates 34% of Tracks Received Daily

THE Business Season reported that Deezer receives more than 50,000 AI-generated tracks daily - more than 34% of the total.

And here's the scary thing: a Deezer/Ipsos survey of 9,000 people showed that 97% got it wrong when trying to differentiate AI-generated music from music created by humans.

Producer Rick Bonadio put it bluntly: music made by AI is not art. But the reality is that consumers can't tell the difference - and that changes everything.

On the bright side? 76% of Brazilians want transparency about AI content, and 65% believe that AI threatens the pay of musicians, calling for copyright protection.

Fake AI Image Cancels Trains in England

And to show that the risks of generative AI are not just theoretical, the BBC reported that a false image generated by AI led to the cancellation of trains for an hour and a half in the north-west of England.

The image showed Carlisle Bridge partially destroyed after a 3.3 magnitude tremor. Network Rail halted services for a safety inspection, resulting in 32 delayed services and significant costs to the taxpayer.

Expert Tony Miles noted that the biggest problem was not the delay itself, but the unnecessary mobilization of checking teams, This can have an impact for days.

What Do These Facts Together Reveal?

When we look at the last 24 hours as a whole, three patterns emerge clearly:

1. Technological Sovereignty Is a Matter of National Security

Trump is treating AI as a strategic priority of state - centralizing regulation, negotiating billion-dollar exports with US-friendly conditions, and positioning the technology as a matter of “world domination”.

China, for its part, continues to launch globally competitive models, bypassing sanctions and investing heavily in infrastructure.

What about Brazil? It is debating copyright - which is important, but cannot be the only strategic agenda.

2. Regulatory Fragmentation Is a Real Risk

Trump has a point (albeit a controversial one): 50 different state regulations can really put the brakes on the industry. This doesn't mean that authoritarian centralization is the answer, but clarity, predictability and harmonization are urgently needed.

In Brazil, we need a regulatory framework that protects rights without hindering innovation. And that's what we need now - not in three years.

3. Trust is the Scarcest Asset

Workers who train AI warning family members not to use it. Musicians watching 34% from the market flooded with indistinguishable synthetic content. Trains being canceled by fake images.

Technology has advanced faster than our ability to evaluate, audit and trust it. E trust does not automatically scale with computing power.

What does this mean for Brazilian companies and professionals?

If you work with AI, lead digital transformation projects or are simply trying to understand how your company should position itself in this scenario, here are some practical thoughts:

For companies

• Attention to ongoing regulation: If you plan to train proprietary models with Brazilian data, keep a close eye on the legislative discussions on copyright and TDM (Text and Data Mining)
• Consider transparent partnerships: Tools such as ProRata.AI and clear licensing models can be competitive differentiators
• Diversify suppliers: Don't rely exclusively on a single generative AI provider - especially if they are subject to geopolitical restrictions
• Invest in governance: Having clear auditing, explainability and security processes is no longer optional - it's a minimum requirement

For Professionals

• Develop a critical technical sense: Knowing how to use ChatGPT is different from understanding its limitations, biases and risks
• Specialize in regulated niches: Areas such as health, legal and finance will require AI with specific compliance - be the person who understands it
• Connect technology with impact: The most value-added AI projects in the coming years will be those that solve real problems responsibly

My Vision: We Need Strategic Pragmatism

I've been working with governments, companies and innovation ecosystems for years, and I can safely say: Brazil cannot afford to be the most restrictive country in the world in terms of AI.

This doesn't mean giving up copyright, privacy or security. It means being strategic, pragmatic and fast.

We need regulation that:

• Protect Brazilian creators without making national startups unviable
• Allow responsible experimentation in controlled environments (regulatory sandboxes)
• Encourage local infrastructure (data centers, GPUs, talent)
• Establish mandatory transparency on training data and models
• Create fair and technically feasible remuneration mechanisms

And, above all, we need speed. While the US centralizes regulation and negotiates billions with China, and China launches models that reach 10 million users in a week, Brazil cannot spend years debating the same law.

Conclusion: The Window of Opportunity is Closing

The last 24 hours have shown us that the race for AI is not just technological - it's geopolitical, regulatory and, fundamentally, about who will have sovereignty to decide their own digital future.

Brazil has talent, a market and potential. But talent without infrastructure doesn't scale. A market without regulatory clarity won't attract investment. And potential without strategy cannot be realized.

I'm not asking us to copy Trump or Xi Jinping. I'm asking us to have the the courage to be pragmatic, the speed to be relevant and the wisdom to be inclusive.

Because if we miss this window - and it's closing - we won't just be losing R$ 21 billion. We'll be losing an entire generation of technological capacity.

In my mentoring and consulting work, I help executives, companies and governments navigate exactly these strategic crossroads - where technology, regulation and impact meet. If your organization is trying to understand how to position itself in this rapidly changing landscape, let's talk. Because the decisions we make today will define whether we will be protagonists or spectators in the age of artificial intelligence.

And I don't know about you, but I prefer protagonism.