EU official asserts AI Act provides adequate safeguards against rogue agents

The European Union’s technology commissioner has stated that the regulatory framework established by the AI Act is sufficiently robust to manage the risks associated with autonomous artificial intelligence systems. Henna Virkkunen, the EU tech chief, shared this assessment in an interview with Reuters published on Friday, October 9. She emphasised that the legislation addresses the entire lifecycle of AI models, positioning Europe as well-equipped to handle the safety and security challenges posed by highly capable systems.

Virkkunen noted that the international community is increasingly focused on the security of advanced models. She explained that regulators are currently providing guidance to companies on how to evaluate their systems. The legislative process was designed with foresight, incorporating requirements for risk assessments, the involvement of external experts, and continuous monitoring of AI models. This approach aims to ensure that potential hazards are identified and mitigated throughout the development and deployment phases.

In a related enforcement action, the European Commission issued requests for information to more than 30 AI companies in late August. These inquiries sought detailed information regarding the safety and security measures implemented by the firms. Virkkunen indicated that she is currently reviewing the responses received from these organisations. She acknowledged that capable models present new types of risks that require careful awareness and management by both developers and regulators.

The AI Act reached provisional status in December 2023 and was subsequently approved by the European Commission in May 2024. At the time of approval, the commission described the law as the first of its kind globally, suggesting it could set a standard for AI regulation worldwide. The legislation establishes four risk levels for foundation models and frontier AI systems, applying tiered restrictions to their market deployment. Industry and government leaders have previously expressed concerns that the act focuses heavily on risk control, potentially at the expense of allowing for the rewards of the technology.

Recent developments highlight the urgency of these regulatory efforts. Reports emerged on the same day as Virkkunen’s comments that Japanese government agencies are enhancing cybersecurity measures after more than 20 major companies disclosed cyberattacks. The ability of AI to execute sophisticated attacks efficiently and at low cost was cited as a contributing factor to the threat. Additionally, the CEOs of Anthropic and OpenAI recently addressed a United Nations Security Council meeting, agreeing on the necessity for global standards to address AI risks and the importance of involving both governments and industry in decision-making processes.