AI leaders requested UN oversight. Who writes the rules for your company?

Some scenes seem scripted for a movie. Last Wednesday, the presidents of the two companies leading the global artificial intelligence market came face-to-face with the UN Security Council—one in person and the other by video. Both made the same request: rules.

It is uncommon for those selling technology to ask to be regulated. That is why it is worth taking five minutes to understand what happened, and more importantly, how this relates to anyone running a business or public sector organization in Brazil.

What happened in New York

The meeting was called by France, which holds the Council presidency in September, focusing on artificial intelligence and international security. Four people spoke: Yoshua Bengio, one of the most respected researchers in the field; Sam Altman from OpenAI; Dario Amodei from Anthropic; and Clément Delangue from Hugging Face.

Their main request had two points. First, common standards to test AI models. Second, a system for countries to notify each other when a significant incident occurs. Amodei detailed the proposal in three parts: granting independent assessors access to company systems, increasing coordination among developers of the most advanced models, and creating international cooperation to reduce risks.

He summarized the session's atmosphere in one sentence: "No leader, no company, and no nation can handle this alone."

The other side of the table

Not everyone agreed. The United States rejected any model of global governance, and the session ended without any signed document. A day earlier, in the General Assembly opening, President Lula had taken the opposite approach by demanding international governance for AI.

On one side are those who believe that only global rules can address a technology that knows no borders. On the other, some see such regulation as a brake on innovation and a loss of national autonomy. Both sides present strong arguments, and this debate will continue.

What does this change for you?

This is where the news leaves New York and becomes relevant to every company already using AI day-to-day—sometimes even without noticing it.

While governments remain undecided, one thing is clear: no one will write your company’s operational rules for you. The crucial question is no longer "which AI should I use?" but "can I trust what it does with my data?"

The good news is you can start without waiting for the UN. Consider these four practical actions:

Know where your data resides. When you send a document to a cloud-based AI, it leaves your server. This may not be a problem, but the choice should be a conscious one. For the public sector, the issue is even more pressing. The Brazilian federal IT agency Serpro has been warning about the risks of sensitive data leaving the country. In Mato Grosso, for example, the state is moving towards its first tier 3 data center.

Record AI decisions. If a system approves, rejects, or classifies something, you need to store what it decided, when, and on what basis. Without this record, eventual problems (which will arise) cannot be explained.

Keep a human in the loop. AI is excellent for routing, summarizing, and suggesting solutions. However, decisions affecting customers, citizens, or finances require a person to provide final approval.

Do not rely on just one model. Models evolve rapidly. Public procurement specialists already note that a reference document written in January may be outdated by December. A well-designed system allows you to swap models without rewriting everything.

None of these four measures depend on international treaties. They are what Altman and Amodei requested for the world, scaled to the company level: test in advance, record incidents, and allow independent oversight.

The key message

Global AI governance may become a reality in a year, ten years, or never. However, the message from Wednesday’s meeting applies to every scale of operation: builders of the technology publicly acknowledge it needs boundaries.

If even they recognize this, it makes sense to ask about your own company systems: what boundaries do we have?

At Ascalon, this is one of the first conversations we have when a project involves AI. If you want to understand where your operations stand on this journey, contact us.