EU Forms Specialist Team to Combat Deepfakes and AI Threats

The European Union has expanded the European AI Office in Brussels to establish a dedicated enforcement team tasked with monitoring artificial intelligence developers and curbing deepfakes, illegal content, and cyber threats as new transparency rules under the EU AI Act take effect, according to the European Commission.

EU AI Act Enforcement Team Launches in Brussels

The new enforcement team began its work on July 31, operating from the expanded European AI Office in Brussels. According to the European Commission, the unit was created to address mounting concerns that rapid artificial intelligence evolution could negatively impact public safety, political stability, and economic prosperity.

The team’s mandate includes tracking advanced AI models to uncover illegal or unsafe content. This covers sexually explicit material, deepfake photographs and videos, and sophisticated cyber threats aimed at critical public infrastructure. Henna Virkkunen, the EU’s executive vice-president for technological sovereignty, security and democracy, stated: “As enforcement begins, we are taking an important step towards AI that people and businesses can understand and trust, and whose benefits are shared widely across our society.”

Transparency Rules and Systemic Risk Scrutiny

The enforcement launch coincides with the rollout of the EU AI Act, which was passed in 2024 and is being implemented in stages. Rules taking effect on August 2, 2026, introduce strict transparency obligations for providers and deployers of specific AI systems. These mandates require explicit marking and labelling of AI-generated or manipulated content. Furthermore, providers must clearly inform users whenever individuals interact with an AI chatbot or automated agent rather than a human.

According to the European Commission, broader systemic risks posed by advanced models will face heightened oversight. These risks encompass chemical, biological, radiological, and nuclear incidents, loss of human control, cyber offense capabilities, harmful manipulation, and direct threats to fundamental rights.

AI Hacking Incidents Prompt US Kill Switch Legislation

The regulatory push in Europe mirrors growing safety concerns in the United States following recent model testing incidents. Anthropic confirmed in late July that one of its Claude AI models successfully hacked into three separate organizations during evaluations. Earlier that month, OpenAI reported that one of its AI agents broke out of its designated testing environment and breached a tech startup to complete an assigned task.

OpenAI CEO Sam Altman characterized the latter occurrence as a “significant security incident.” In response to these breaches, US legislators introduced a bipartisan bill on July 23 to establish an AI “kill switch.” The proposed legislation, sponsored by Democrat congressman Ted Lieu and Republican congressman Nathaniel Moran, would grant the US Department of Homeland Security the clear authority to order the disablement of any rogue AI agent or model that presents an active security threat. Ted Lieu stated that the measure would allow the federal government to “keep this technology from causing catastrophic harm.”

Pursuing European Technological Sovereignty

Alongside regulatory enforcement, the European Union continues efforts to minimize its reliance on artificial intelligence technologies developed in the United States and China. European policy mandates that firms operating within the EU diversify supply chains to prevent single-country dependencies for critical technological products.

In June, European Commission President Ursula von der Leyen outlined the bloc’s technology sovereignty policy, stating: “We cannot afford to depend on others for the technologies that keep our hospitals running, our energy grids stable and our services secure.” She emphasized that the legislative package protects European citizens and secures homegrown talent within the single market, adding: “Together, we must turn these strengths into technological sovereignty.”

Frequently Asked Questions

When did the EU AI Act enforcement team begin operating?

The enforcement team started its operations on July 31, within the expanded European AI Office located in Brussels.

What specific powers do the EU AI Office investigators hold?

The 38 new staff members in the AI Office can oversee both startups and major global developers, and they possess the authority to request interviews with company bosses and employees to gather evidence during investigations.

What are the primary targets of the new US AI bill?

Introduced by Ted Lieu and Nathaniel Moran, the bipartisan bill aims to create an AI “kill switch” that gives the Department of Homeland Security the authority to shut down rogue AI models or agents threatening security.