The EU AI act: between the rock of product safety and the hard place of fundamental rights

“In contrast, the United States has adopted a more decentralized, innovation-oriented strategy, with sector-specific guidelines rather than unified federal regulation (Allen et al., 2025; Almada & Petit, 2025; Sloane & Wüllhorst, 2025).”

“One possible reason mentioned for companies’ caution is that current RAG systems do not yet fully comply with the stringent requirements of the EU AI Act (Almada and Petit, 2025).”

“Some scholars, such as Almada and Petit (2025), have proposed introducing a “ strict liability ” mechanism for high-risk AI systems.”

“…liability (Trump Administration 2025), or China, where governance is framed through industrial policy and targeted regulatory interventions in prominent domains (Chen and Xu 2025), the EU has opted for a comprehensive, fundamental rights-based, and risk-oriented framework (Almada and Petit 2025).”

“Like in the EU’s artificial intelligence regulation, there is thus a probabilistic association between a purpose and a risk [2].”

“The AI Act’s additional specific focus on the protection of fundamental rights further complicates matters: While a standard allows for a clear distinction — compliant or non-compliant — the protection of fundamental rights usually requires context-specific proportionality assessments [2].”

“The AI Act (Almada and Petit 2025) classifies AI systems into minimal, limited, high, and unacceptable risk categories, with the strictest obligations on high-risk systems, and reflects the seven Trustworthy AI requirements articulated by the High-Level Expert Group on Artificial Intelligence (AI…”

“2 Indeed, compliance with new regulations can be overwhelming, especially for smaller companies or startups.”

“AI systems posing transparency risks are subject to disclosure obligations, whereas low-risk or de minismis AI systems are not covered by AI Act (Almada & Petit, 2025; Razquin, 2024).”

“Yet it does not address the broader governance problems that motivated the AI Act, including technical robustness, post-market monitoring, value-chain responsibility, and system-level risk management (Almada and Petit, 2025, pp. 85–87).”

“Almada [1] offers a detailed legal examination of how product safety and fundamental rights protections are jointly pursued in the AIA, explicating the dual objectives that underpin the high-risk regime and the interpretive tensions this duality can create.”

“…obligations may produce functionally different regulatory outcomes, making the implementation phase an integral component of harmonisation itself (Almada and Petit, 2025), whose effectiveness (effet utile) depends upon sufficiently coherent administrative and technical practices across Member…”

“36 This is particularly important in the healthcare context, where AI systems directly affect legally protected interests such as the right to health, non-discrimination, privacy and human dignity.”

“Whereas the earlier product safety regulations sought to protect against risks to health and safety, the AI Act also addresses risks to fundamental rights (Almada and Petit 2025).”

“The latter demands a maximisation of the realisation of fundamental rights (Almada & Petit 2025), rather than the regulatory balancing of fundamental rights protection against the private costs of risk prevention and mitigation or the loss of private profits.”

“…three research questions to four source domains: (i) Regulation (EU) 2024/1689 and the legal scholarship interpreting its governance architecture [13–22]; (ii) EU statutory-audit law [56,57]; (iii) IAASB pronouncements on quality management, engagement quality, and audit evidence [36]; and (iv)…”

“Owing to the rapid advancement of AI technologies, regulators often struggle to keep pace with new developments, and existing frameworks may quickly become outdated and/or fail to address pertinent issues [2].”

“Third, one can also argue that the EU AI Act can be seen, at least partly, as product safety law (Almada & Petit, 2025).”