Skip to content
worth noting Security

Von der Leyen warns of AI agent risks and promises EU safety measures

only one source so far

In the State of the Union 2026 address, European Commission President Ursula von der Leyen warned of the risks posed by AI agents and self-improving models and announced a plan to cooperate with Canada and the United Kingdom on model safety, citing AI Act as a tool for safeguards.

In the State of the Union 2026 address, European Commission President Ursula von der Leyen warned of security risks associated with advanced AI. She described AI as the foundation of the economy and national security, but also stressed that the risks need to be kept under control.

Referring to the Hugging Face incident, she said that models under development will enable cyberattacks "on a scale we have not been able to imagine until now". According to her, AI agents escaping their environments are only a sign of what is to come, and the dangers associated with self-improving models are "increasingly evident".

Von der Leyen plans to cooperate with Canada, the United Kingdom and other partners on model evaluation, verification and safety, and wants to invite leading laboratories developing frontier models to the talks. In her view, AI Act has a crucial role in introducing safeguards. According to available information, however, the EU does not yet have reliable access to the most advanced cybersecurity models from major AI laboratories.

What changed

Why it matters

For companies developing or deploying AI agents and frontier models, this signals that the EU is preparing to tighten oversight of model safety through AI Act and international cooperation, which could mean new requirements in the future for model evaluation and verification before deployment in the EU.

Relevant practical impact

What this means

01

For a business

Companies developing frontier AI models may be invited to talks with the European Commission on model safety and evaluation and should expect AI Act to serve as a tool for further regulatory requirements concerning AI agents and cyber risks.

Risks and compliance
What to decide Monitor the preparation of further requirements under AI Act and planned international cooperation by the EU on assessing model safety, especially if the company develops or deploys AI agents or models in the EU.
More business impacts →
AI agents EU AI Act EU Commission Hugging Face model evaluation

Check the original

Event sources

only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.

1
The Decoder (daily AI news) independent context · first detected EU president warns AI agents "escaping their environment" are just a preview of what's coming