The EU AI Act and USA AI.gov Action Plan: A Legal Comparison
Artificial Intelligence (AI) governance is developing along two main paths. The U.S. AI.gov Action Plan and the EU AI Act take very different approaches. The U.S. focuses on innovation, market forces, private sector leadership, and national security, relying on current laws and executive guidance instead of detailed bans. The EU uses strict, binding rules that prioritise fundamental rights and set precise requirements based on risk. These differences shape the governance models discussed in this analysis.
The United States encourages AI development by lowering regulatory barriers and using existing laws to address challenges. In contrast, the European Union has created a broad legal framework that also applies to companies outside its borders.
The U.S. AI.gov Action Plan: A Strategic Framework for Innovation and Security
The U.S. AI.gov Action Plan sets out how the country will approach artificial intelligence. It aims to speed up innovation, strengthen AI infrastructure, and lead in international AI diplomacy and security. The plan relies on current laws and encourages private sector growth instead of adding new, strict regulations.
Core Legal Principles and Regulatory Intent
A central tenet of the U.S. plan is fostering private sector-led innovation by reducing regulatory burdens. The action plan states its intent to “remove red tape and onerous regulation”. This shows a preference for market-driven development and a belief that innovation flourishes with minimal government oversight. The plan assumes existing legal frameworks, industry self-governance, and voluntary standards are enough to manage emerging risks without stifling technological advancement.
The plan also demonstrates a strong commitment to protecting American values, free speech, and intellectual property. It highlights the importance of ensuring “Frontier AI Protects Free Speech and American Values” and “Protecting Commercial and Government AI Innovations“. This focus suggests an alignment of AI development with fundamental constitutional principles and the safeguarding of economic interests. Such protections are anticipated to be enforced primarily through existing intellectual property laws, anti-discrimination statutes, and new executive guidance, rather than through a dedicated, overarching AI statute.
A strategic focus on national security, secure-by-design AI, and incident response is another critical component. The plan underscores a proactive stance on managing geopolitical and operational risks posed by advanced AI, emphasising “Ensuring that the U.S. Government is at the Forefront of Evaluating National Security Risks in Frontier Models,” “Promoting Secure-By-Design AI Technologies and Applications,” and “Promoting Mature Federal Capacity for AI Incident Response“. This indicates a strong governmental role in defence and critical infrastructure applications, focusing on developing robust capabilities to respond to AI-related incidents or misuse. This approach reflects a recognition that AI systems, particularly frontier models, can pose significant national security challenges that require a coordinated governmental response.
The plan addresses the misuse of AI in legal contexts, particularly synthetic media like deepfakes and manipulated evidence. This requires adaptations to evidentiary rules, forensic capabilities, and legal precedents within the existing judicial system.
Implicit Governance Structures and Implementation Approach
The U.S. AI.gov Action Plan positions the Federal government as the primary entity responsible for creating the conditions where private sector-led innovation can flourish. This suggests an enabling, rather than strictly controlling, role, guiding the overall environment for AI development through policy and strategic initiatives. Specific departmental involvement is noted, with the Department of Defence (DoD) explicitly mentioned for driving AI adoption, indicating its crucial role in military AI applications and related governance. While other departments are not detailed in the provided information, the nature of the plan implies a distributed approach across various agencies, each addressing AI within its existing mandate.
Furthermore, the U.S. aims to “Counter Chinese Influence in International Governance Bodies“, signifying an active intent to shape global AI norms and standards through diplomatic engagement and multilateral forums rather than solely through domestic regulation. This highlights a preference for international influence and norm-setting as a key component of its governance strategy.
A significant characteristic of the U.S. approach is the absence of a single, comprehensive statutory framework. The document explicitly states that it does not expressly detail specific legal frameworks, agencies, or detailed regulatory bodies that will implement these principles, but instead outlines the strategic direction. This indicates a reliance on existing laws, executive orders, agency guidance, and sector-specific regulations, rather than a monolithic, overarching AI Act.
Using a policy statement instead of a complete legal framework can lead to uncertainty and confusion. This flexible approach may help drive innovation, but it can also create different rules across agencies, making it harder and less predictable for businesses to follow the law.
The plan highlights a tension between protecting intellectual property and promoting open innovation. Balancing these priorities will likely lead to future legal challenges, particularly regarding the protection of proprietary models and the liability of open-source AI, especially for advanced systems with national security implications. Ongoing legal interpretation and potentially new frameworks will be needed as these issues evolve.
Mentioning ‘Combatting Synthetic Media in the Legal System’ and ‘Promoting Mature Federal Capacity for AI Incident Response’ shows the U.S. anticipates legal system challenges.
The EU AI Act: A Comprehensive, Risk-Based Regulatory Framework
The European Union’s Artificial Intelligence Act (AI Act), formally Regulation (EU) 2024/1689, stands as the world’s first comprehensive legal framework on AI, setting a global precedent for AI governance. This pioneering regulation came into force on August 1, 2024, with provisions gradually becoming operational over the following 6 to 36 months.
Foundational Principles and Objectives
The core objective of the AI Act is to ensure that “Europeans can trust what AI has to offer” by addressing risks and ensuring AI systems “respect fundamental rights, safety, and ethical principles“. This rights-centric and ethics-driven approach is a cornerstone of EU regulation, consistent with its broader data protection framework, the General Data Protection Regulation (GDPR). The Act aims to foster trustworthy AI in Europe and beyond by establishing a common regulatory and legal framework for AI within the EU.
A critical aspect of the AI Act is its extraterritorial application. The Act explicitly applies to “providers and developers of AI systems that are marketed or used within the EU (including free-to-use AI technology), regardless of whether those providers or developers are established in the EU or another country“. This broad reach, similar to GDPR, means that global companies must comply with the Act’s provisions if they operate in the EU market, effectively extending the Act’s influence beyond EU borders.
This extraterritoriality, combined with its pioneering comprehensiveness, creates a powerful regulatory export mechanism, often referred to as the “Brussels Effect.” Companies globally, seeking access to the significant EU market, will likely adapt their AI systems to meet EU standards, effectively making the AI Act a de facto global benchmark. As a result, EU values and rules are likely to shape AI development worldwide, setting international standards and possibly encouraging other regions to follow similar regulations.
The phased implementation of the Act allows stakeholders to adapt and comply as provisions become operational over 6 to 36 months.
The Risk-Based Classification System and Associated Obligations
The AI Act employs a “proportionate risk-based approach,” classifying AI systems into four levels: Unacceptable, High, Limited, and Minimal risk, with an additional category for General-Purpose AI (GPAI) and GPAI with systemic risk (GPAISR).
Unacceptable Risk – Banned Practices: AI applications that are considered a clear threat to the safety, livelihoods, and rights of people are banned. These prohibitions are exhaustive and include:
By clearly banning specific “unacceptable risk” AI practices, the EU demonstrates its commitment to protecting fundamental rights and society, even if it means giving up some uses of technology. This is different from the U.S. approach, which values free speech and American values but does not list specific AI bans, instead relying on existing laws to address problems as they arise.
High-Risk AI Systems – Strict Obligations: These systems are expected to pose “significant threats to health, safety, or the fundamental rights of persons”. They are subject to extensive ex-ante (before market placement) and ex-post (throughout their lifecycle) obligations. Examples of high-risk use cases include AI in health, education, recruitment, critical infrastructure, law enforcement, and justice.
Minimal Risk AI Systems: These AI systems typically operate without predefined rules, but companies are encouraged to establish their own guidelines.
General-Purpose AI (GPAI) and GPAI with Systemic Risk (GPAISR): A specific category with tailored transparency and additional strict obligations for GPAISR, defined by their “particularly high-impact capabilities due to their technical strength or reach in the market.“
Governance, Enforcement, and Compliance Mechanisms
The newly created EU AI Office will play a central role in overseeing the implementation and enforcement of the AI Act. 3 (“Decision – 2024/1459”, 2024). This centralized oversight body is crucial for ensuring consistent application across member states.
The Act prescribes “hefty penalties” for non-compliance, ranging from €35 million or 7% of global revenue to €7.5 million or 1.5% of revenue, depending on the infringement and company size. These substantial penalties underscore the seriousness of compliance and serve as a significant deterrent.
High-risk AI systems must undergo rigorous conformity assessments to demonstrate compliance with the Act’s requirements, with specific procedures depending on the application of harmonized standards or common specifications. Notably, substantial modifications to high-risk AI systems necessitate a new conformity assessment, ensuring ongoing compliance throughout the system’s lifecycle. Market surveillance authorities and notified bodies play a crucial role in overseeing compliance and conducting evaluations, with specific provisions for AI systems intended for use by law enforcement, immigration, or asylum authorities.
Although the Act is strict, it also allows for “AI Regulatory Sandboxes” and real-world testing of high-risk AI systems. This shows that lawmakers recognize AI is changing quickly and that too much regulation could slow down practical innovation. These sandboxes enable companies, particularly smaller ones, to test and refine their AI under supervision before facing all compliance requirements. This approach tries to balance strong rules with support for innovation.
The Act also clarifies that existing EU law on the protection of personal data, privacy, and confidentiality (e.g., GDPR) applies to the collection and use of any such information for AI-based technologies, ensuring a comprehensive data governance framework.
Despite its stated goals of protecting fundamental rights, the Act has drawn criticism. For instance, La Quadrature du Net (LQDN) described the AI Act as “tailor-made for the tech industry, European police forces as well as other large bureaucracies eager to automate social control” and “largely incapable of standing in the way of the social [control]“. It implies that even a comprehensive legal framework might face challenges in fully safeguarding against state or corporate overreach, particularly in areas like surveillance and data exploitation, and that ongoing vigilance and potential future amendments or challenges may be necessary.
Conclusion
The U.S. and EU take fundamentally different paths for AI regulation. The U.S. favours innovation and security, with an adaptive, flexible approach. The EU stresses rights, risk mitigation, and trust through a uniform, binding framework. Each approach reflects distinctive legal priorities and risk management philosophies.
The U.S. AI.gov Action Plan lacks legal force, relying on current agencies and laws. The EU AI Act is a binding regulation, enforced by a new EU AI Office, with strict penalties and unified standards for the internal market. U.S. policy primarily addresses domestic concerns, while the EU AI Act has a broad extraterritorial scope, influencing global norms. The U.S. takes a case-by-case approach to risks, whereas the EU imposes mandatory bans and strong obligations based on risk tier. The U.S. relies on a decentralised, flexible enforcement model, which risks fragmentation. The EU centralises oversight and sets high compliance standards. Each side views systemic risk through its own lens: U.S. security versus EU rights and safety.
These divergent models challenge global businesses navigating compliance and operational complexity. The EU’s binding standards and extraterritorial effect give it first-mover influence, likely shaping future global AI regulation.


