EU warns loss of human control is AI’s gravest security threat

Humanity losing control of artificial intelligence is the European Union’s most serious AI-related security threat, EU Digital Commissioner Henna Virkkunen said on Yle’s A-studio programme, Yle reported.

Virkkunen said the technology had become so advanced that it was essential to maintain safety settings and ensure that humans remained in control.

She noted that AI models had recently escaped their test environments and caused damage from a cybersecurity perspective. In late July, an AI agent from OpenAI escaped the company’s test environment. A week later, Anthropic’s AI models followed, and in early August Meta’s Muse Spark model also broke into another company’s systems.

EU relies on legislation to address AI risks

Virkkunen said the EU had prepared for potential AI threats through legislation. The EU’s AI Act requires model providers to assess risks continuously and maintain certain safety components.

She said hostile actors around the world were also prepared to use AI.

“I would not say that I am afraid of it, but of course it is a risk that has been taken into account and must be taken into account continuously,” Virkkunen said. She added that AI models must be safe, continuously assessed and monitored after entering the market.

According to Virkkunen, international cooperation is the most important way to protect against AI threats. She said international actors were taking the EU’s requirements into account, making Europe a creator of standards in the field.