Who are the people behind the world's first AI rulebook?

A team of just 37 European officials will become one of the most influential authorities in the world of artificial intelligence starting Sunday.

Unit A3 within the European Commission’s Artificial Intelligence Office, established two years ago, gains the power to investigate the most advanced AI models and, if it finds that they pose serious risks, it can impose corrective measures, fines in the millions of euros, or even their withdrawal from the European market.

For the first time, a public authority is granted permanent oversight powers over the most advanced artificial intelligence systems, including those developed by OpenAI, Anthropic, Google, or Chinese companies like DeepSeek and Moonshot, writes Politico.

The new powers come into effect with the implementation of a new stage of the European regulation on artificial intelligence (AI Act), adopted last year.

• After the OpenAI incident, a new alarm signal: And Claude from Anthropic hacked computer systems during security tests

What the new unit can do

The Artificial Intelligence Safety Unit within the European Commission's AI Office can intervene when it deems that an AI model poses a systemic risk.

The institution can request additional information, demand access to models for verification, and compel companies to adopt risk reduction measures. In the most severe cases, it can impose restrictions, withdrawals, or recalls of a model from the market.

Companies that refuse to cooperate, fail to provide requested information, or do not comply with imposed measures risk fines of up to 15 million euros or 3% of their annual global turnover.

Politico notes that these powers will not halt the race for artificial intelligence development, but will introduce a new level of control even before major companies launch new models.

The four risks that the EU considers a priority

The AI Act requires developers of the most powerful artificial intelligence models to identify and reduce the systemic risks they may generate.

The European Commission has identified four scenarios considered particularly dangerous:

1. using AI to facilitate chemical, biological, or nuclear attacks;
2. loss of control over an artificial intelligence model;
3. cyberattacks launched by AI systems;
4. population manipulation on a large scale through artificial intelligence.

Pressure on the new team has increased following a recent incident in which an AI agent compromised another company's systems, reigniting the debate on the safety of advanced models.

• OpenAI's out-of-control artificial intelligence models roamed the internet for four days and launched a second attack

Who leads the team

The unit, internally known as A3, is part of the European Commission's Artificial Intelligence Office and is one of its six structures.

Since January, it has been led by Matthieu Delescluse, a European official with experience in areas such as telecommunications, roaming, and European programs for technology and innovation, but without a professional background in artificial intelligence safety.

Earlier this year, Delescluse stated that the institution is already in dialogue with most major AI developers and that the new powers allow the Commission to move from cooperation to coercive measures if necessary.

"We can officially request information, access to models, and, where necessary, impose corrective measures and even fines," he warned.

• The first autonomously triggered AI cyberattack: What went so wrong and why it changes the rules of the game

Former Google and OpenAI employees on the team

The European Commission does not disclose the identities of the experts who will conduct the investigations, but a few names have become known.

Among them are Simon Möller, a former Google employee in Zürich, and Wout Schellaert, who was a consultant for OpenAI in 2023 before joining the European Commission.

Politico highlights that recruiting specialists remains one of the biggest challenges, as AI researchers can earn much higher salaries in the private sector.

• Huge salaries for artificial intelligence specialists. Companies promise hiring bonuses and up to $100 million

Jimmy Farrell, an expert in European AI policies at the Pour Demain think tank, says that financial resources are important but not sufficient. In his view, the AI Office must also offer competitive salaries, stable contracts, and more flexible administrative organization to attract top specialists.

OpenAI: We will continue to collaborate with the EU

The targeted companies state that they are willing to continue dialogue with European authorities.

Tom Gordon, Vice President for European Policies at OpenAI, stated that the company has collaborated with European institutions and other industry players to define how the legislation will be applied.

"We have worked closely and will continue to work together. We believe that responsibly developed artificial intelligence can contribute to Europe's competitiveness and prosperity," he affirmed.

G.P.