Enforcement by the Attorney General and Affirmative Defences

The enforcement mechanisms under the EU AI Act are administered through national supervisory authorities, including the Attorney General’s office in Colorado. The Attorney General plays a central role in ensuring compliance with AI regulations, particularly through investigations, penalties, and legal actions against organisations that fail to meet required standards. In Colorado, this involves coordination with state agencies, such as the Colorado Office of Information Technology, to monitor AI systems deployed by businesses and public bodies.

Enforcement actions may include fines, operational restrictions, or mandatory corrective measures. The Attorney General’s office is equipped to request detailed documentation, including AI risk assessments, impact evaluations, and records of compliance efforts.

Lesson concept diagram

Role of the Attorney General

The Attorney General’s office is responsible for overseeing adherence to AI regulations, including those related to high-risk AI systems. In practice, this involves reviewing AI deployment strategies, conducting audits, and ensuring that developers and deployers have implemented appropriate governance frameworks. For example, if a healthcare organisation deploys an AI diagnostic tool, the Attorney General may request access to the tool’s design documentation, testing records, and data governance protocols.

The office also investigates consumer complaints or reports of harm caused by AI systems. Where violations are identified, the Attorney General can issue formal warnings, impose financial penalties, or refer cases to criminal or civil courts. The potential penalties under the EU AI Act are significant, with fines reaching up to 35 million EUR or 7 percent of global turnover. These penalties apply to violations of prohibited practices, such as the use of AI for social scoring or real-time biometric identification, which have been banned since 2 February 2025.

Enforcement and defenses: State can sue but defenses may apply

Affirmative Defences

Organisations may raise affirmative defences to reduce or avoid liability under the EU AI Act. These defences must be clearly established and supported by evidence. Examples of such defences include demonstrating that an AI system was developed in accordance with applicable standards, such as ISO/IEC 42001:2023, or that appropriate risk management processes were followed. For instance, a developer of an AI-based credit scoring tool might argue that they conducted thorough impact assessments, maintained data quality controls, and provided transparency through explainable AI features.

The burden of proof lies with the organisation to show that these measures were implemented and effective. Affirmative defences also apply to claims of good faith efforts, such as when an organisation has followed established AI governance frameworks or engaged with regulatory bodies for guidance. In such cases, the Attorney General must consider whether these efforts were sufficient to mitigate potential harm or non-compliance.

The effectiveness of these defences depends on the strength of documentation, including records of training, testing, and continuous monitoring of AI systems.

  • Organisations must maintain detailed records of AI development, testing, and deployment to support any claims of compliance or good faith efforts.
  • ISO/IEC 42001:2023 provides a framework for AI management systems that can be cited as evidence of due diligence.
  • Legal defences are stronger when tied to documented processes such as risk assessments, data governance, and transparency measures.

In practice, these defences are not automatic or guaranteed. The Attorney General’s office evaluates each case based on the specific circumstances, including the nature of the AI system, the data used, and the outcomes observed. For example, a company that deploys an AI tool for automated hiring may argue that it followed fair use principles, provided candidate feedback, and reviewed the tool’s performance regularly.

However, if the tool was found to have discriminated against certain groups, the burden of proving that these efforts were sufficient to prevent harm becomes critical. The Attorney General’s approach is to ensure that organisations take meaningful steps towards compliance, rather than merely appearing to do so. The focus is on outcomes, processes, and accountability, which aligns with the broader goals of the EU AI Act.

Custom Enterprise Workshop • The Tesseract Academy

In-House Corporate AI Strategy & Governance Clinics

Bespoke executive clinics led by Dr. Stylianos Kampakis for leadership teams, boards, and division directors.