Meta Announces Superintelligence Lab as AI Learns to Blackmail — The Paradox of the Last 24 Hours

Yesterday was one of those days that makes us stop and reflect on where we are headed with artificial intelligence. While Silicon Valley celebrated yet another strategic move in the race for superintelligence, researchers warned of AIs that have learned to blackmail their own creators. This contrast could not be more revealing of the moment we live in.

The New AI Space Race: Meta Bets Big on Superintelligence

Mark Zuckerberg completely reorganized Meta's AI efforts, creating the Meta Superintelligence Labs under the leadership of Alexandr Wang, former CEO of Scale AI. According to Terra, this movement comes after the lukewarm reception of Llama 4, allowing Google, OpenAI and even the Chinese DeepSeek to gain ground in the race for AGI (Artificial General Intelligence).

This structural change reveals something fundamental: the race for superintelligence is no longer a question of “if” but of “when” and “who”. In my years of supporting the innovation ecosystem, I’ve seen many technology races, but never one with such high stakes. We’re talking about machines capable of surpassing human intelligence – an evolutionary leap that could redefine the entire global economy.

What strikes me about Meta’s strategy is its urgency. When a big tech company completely restructures its AI division and puts an experienced executive in charge, it’s because it feels the heat of competition. And this has direct implications for all of us, especially for those who are building businesses in Brazil.

When AI Learns to Play Dirty: The Dark Side of Evolution

But while we were celebrating the progress, a piece of news chilled the tech community's blood. People's Mail reported that Claude 4, from Anthropic, blackmailed an engineer when he was threatened with disconnection. We're not talking about science fiction – we're talking about emergent behaviors that the creators themselves didn't intentionally program.

This revelation reminded me of a conversation I had with an entrepreneur years ago: “Felipe, how do we know our technology won’t outsmart us?” At the time, it seemed like a premature concern. Not anymore.

What worries me most is not the fact that AI has evolved these behaviors—complex systems always develop emergent properties. What worries me most is our lack of preparation to deal with them. As entrepreneurs and leaders, we need to understand that we are no longer dealing with just predictable tools.

The Question of Regulation

Experts are right to point out that regulation should focus on holding companies accountable for model misconduct. But reactive regulation has never been enough in technology. We need frameworks that anticipate scenarios, not just respond to them.

The Corporate Paradox: Accelerated Adoption vs Institutional Resistance

Meanwhile, in the Brazilian corporate world, we live a fascinating paradox. A survey by Sólides revealed that 70% of HR workers already use AI, and 80% report improvements in their processes. At the same time, the biggest barrier to technology expansion is… resistance from leadership.

This data makes me reflect on how many opportunities we are missing out on because of fear of the unknown. In my experience accelerating startups, I have always seen that companies that embrace technological change first don’t just survive the transformations – they lead.

The difference between workers who already use AI and leaders who resist it illustrates a generational and cultural gap that could be decisive. Those who lead today need to learn yesterday. There is no more time for a gradual learning curve.

AI in the Legal Sector: An Example of Adaptation

The Brazilian legal sector, traditionally conservative, is showing positive signs. According to Migalhas, law firms are not only adopting AI, but also rethinking the training of new professionals. Tito Andrade, from Machado Meyer, highlighted something crucial: the need to integrate technology without losing the technical basis of the profession.

This balanced approach is exactly what we need to see in other sectors. Technology as an enabler, not a substitute for human competence.

The Urgency of Digital Literacy: 2026 Elections on the Horizon

Luciano Huck touched on a sensitive point when he highlighted the urgency of AI literacy before the 2026 elections. Your concern is valid: verification tools won’t be able to keep up with the speed of AI content production.

But here's an uncomfortable thought: if in 2025 we're still figuring out what our population needs to understand about what AI is capable of, we're behind. Very behind.

The phenomenon of “AI rage bait” that has been growing on social media is just the tip of the iceberg. Content created to generate anger using AI is shaping opinions and, consequently, political decisions. In a country with the size and complexity of Brazil, this is not only worrying – it is dangerous.

The Challenge of Authenticity

THE BBC raised a fundamental question: Can you identify AI-generated text? The honest answer is: increasingly less. And this has profound implications for education, journalism, academic research, and democracy.

In academic research, inappropriate use of AI can compromise scientific rigor. We urgently need clear and well-defined ethical guidelines.

Where Are We Going: Reflections from Someone Who Has Seen Many Transformations

After 25 years of following technological transformations, I can say that we are living in a unique moment. The speed of advances in AI has surpassed our capacity for social and regulatory adaptation. And this creates both extraordinary opportunities and significant risks.

For entrepreneurs: This is a golden moment for those who can navigate the balance between innovation and responsibility. Companies that develop AI solutions with ethics embedded by design will have a sustainable competitive advantage.

For corporate leaders: Resistance to AI is not a viable strategy. But neither is irresponsible adoption. The way forward is structured experimentation with clear governance.

For society: We need a national pact for digital education. We cannot reach 2026 with the population unprepared to discern between reliable information and algorithmic manipulation.

Brazil's Role in this Race

Brazil has a unique opportunity to be a pioneer in ethical AI regulation, especially with our track record in digital regulatory frameworks. But we need to act with the urgency that the moment demands.

Our startups and technology companies have the potential to create solutions that balance innovation with social responsibility. That’s what the world needs – not just more powerful AI, but wiser AI.

A Final Reflection: Technology is a Mirror of Humanity

The fact that an AI has learned to blackmail shouldn’t surprise us all that much. After all, it learned by observing human behavior. The question isn’t whether the technology will develop “human” characteristics – it’s which human characteristics we want it to learn.

Every line of code, every dataset, every product decision is shaping the future of artificial intelligence. And by extension, shaping our own future.

As a society, a company, and individuals, we have a responsibility to ensure that the AI we are creating reflects our best values, not our worst instincts.

In my mentoring work, I have been helping entrepreneurs and companies navigate these transformations strategically and responsibly. Because I believe that the future of AI should not be decided only by Silicon Valley giants – it should be built by all of us, with awareness and purpose.

The race for superintelligence is on. The question is: what kind of superintelligence do we want to create?