EU AI Act Seen as De Facto Global Benchmark
The EU AI Act's implementation is expected to set de facto international benchmarks for artificial intelligence governance, according to an EU digital policy analyst. On a recent podcast, the analyst predicted that the legislation's conformance requirements will quickly be adopted by global companies, influencing standards work at bodies like ISO/IEC and the ITU.
- The Act's obligations are being phased in, with the ban on prohibited AI practices effective from February 2025, requirements for high-risk systems starting in August 2026, and full applicability by August 2027. - Providers of "high-risk" AI systems—a category that includes applications in critical infrastructure, medical devices, and employment—must adhere to stringent requirements. These mandates include establishing a continuous risk management system, ensuring high-quality data governance to mitigate bias, maintaining detailed technical documentation, and enabling human oversight. - Compliance with harmonized standards developed by European bodies like CEN and CENELEC will offer a "presumption of conformity" with the AI Act's legal requirements. This elevates the importance of European standard-setting, though these bodies often collaborate with international counterparts like ISO and IEC. - The regulation has a significant extraterritorial reach, applying to any company providing AI systems or services within the EU, regardless of where the company is based. Non-EU providers must appoint an authorized representative within the Union. - Penalties for non-compliance are substantial, with fines reaching up to €35 million or 7% of a company's global annual turnover for violations related to prohibited AI practices. - In contrast to the EU's horizontal, risk-based framework, China has adopted a more state-centric and sector-specific approach to AI regulation. Chinese regulations focus on specific use-cases like generative AI and recommendation algorithms, with an emphasis on aligning technology with national security and social stability goals.
