Control Over AI: European Regulation and Challenges

European regulators and tech companies are actively developing regulations to keep powerful AI systems under control. In Brussels and The Hague, work is underway on new rules, tests, and oversight aimed at limiting risks associated with the use of AI. The European AI Act is being phased in, with specific obligations affecting both government and business. The aim of this regulation is to prevent errors, misuse, and loss of control without stifling innovation.

The rise of generative AI, which generates texts, images, or code, presents various challenges. Companies such as OpenAI, Google, Meta, and Anthropic continually introduce new models applied in various sectors, from customer service to education. Although this technology offers many advantages, there is also the risk that AI systems may make mistakes, leak sensitive information, or be misled. These issues are compounded by the complex accountability structures that arise when different models and fine-tuning are combined by organizations.

The European AI regulation imposes extensive requirements on high-risk AI systems, including obligations for risk assessment, data quality, documentation, and log files. Providers of powerful AI models such as OpenAI and Google must also submit safety tests and information regarding the capabilities and limitations of their systems to the European Commission. This is intended to help ensure safety and the timely reporting of incidents.

The regulation also has implications for government agencies using AI systems. They are required to conduct a fundamental rights assessment before deploying AI. This necessitates training officials in risk recognition and problem reporting. The emphasis is on transparency, for which companies must publish understandable model cards, and the need for strong security measures is essential. Organizations are encouraged to take proactive steps, such as conducting DPIA and FRIA, to minimize the risk of misuse and ensure the safe implementation of AI.

Read the full article from AI Insider.