EU AI Act Transparency Rules Under Threat
Digital rights groups are mobilizing against last-minute amendments to the EU AI Act that could weaken transparency and auditability requirements. European Digital Rights (EDRi) warns that the proposed changes threaten public oversight for high-impact AI systems used in elections and the public sector. The group is urging the European Parliament to reject the proposals.
- The core of the transparency debate revolves around the tension between public accountability and the protection of corporate trade secrets. Proposed amendments risk creating loopholes that would allow developers to shield high-risk systems from audits by claiming their methods are proprietary, undermining the Act's goal of ensuring trust and safety. - Under the Act, deployers of high-risk AI systems—particularly public bodies or private entities providing public services—are required to conduct a Fundamental Rights Impact Assessment (FRIA) before putting a system into operation. This assessment must describe the system's processes, the people likely to be affected, specific risks of harm, and the human oversight measures in place. - The political agreement on the AI Act was reached in December 2023 after marathon "trilogue" negotiations between the European Commission, the Council of the EU, and the European Parliament. Contentious issues included exemptions for law enforcement, the regulation of general-purpose AI models, and the classification of high-risk systems. - Article 13 of the Act specifically mandates that high-risk AI systems be designed for sufficient transparency, enabling users to interpret the system's output and use it appropriately. This includes providing clear instructions for use, detailing the system's capabilities, limitations, and foreseeable risks. - For generative AI, Article 50 requires that AI-generated content like "deepfakes" must be labeled as artificially created or manipulated. Similarly, users must be informed when they are interacting with an AI system like a chatbot, unless it is obvious. - The European AI Office is tasked with developing templates and codes of practice to help companies comply with transparency obligations, but the practical, detailed interpretation of what constitutes sufficient transparency will be shaped over time as the Act is implemented. - Non-compliance with the AI Act can lead to significant penalties, with fines reaching up to 7% of a company's global turnover. Civil society groups have expressed disappointment with the final deal, arguing that self-assessments by developers and exemptions for law enforcement create major loopholes.
