As artificial intelligence, particularly generative AI, becomes embedded across the evidence-generation and medicines-development landscape, regulators and Health Technology Assessment (HTA) bodies are moving quickly to establish guardrails for its responsible use.

Although guidance from NICE, FDA, CDA-AMC, MHRA, and the joint EMA-FDA initiative has emerged from different jurisdictions, a remarkably consistent set of expectations is taking shape. The overarching message is clear: AI can enhance efficiency and innovation, but it must operate within a framework of transparency, validation, accountability, and human oversight.

HTA bodies: Acknowledging AI’s potential, while emphasizing accountability

NICE

NICE’s 2024 Position Statement on AI in evidence generation is among the most influential publications in this area. NICE acknowledges the potential for AI to support systematic reviews, evidence synthesis, real-world evidence generation, and cost-effectiveness modelling. However, it emphasizes that AI should augment rather than replace human expertise. Evidence generated using AI must remain transparent, reproducible, and scientifically robust, with users documenting the tools, methods, and validation procedures employed. NICE also stresses that existing methodological standards continue to apply regardless of whether AI is used. This position has also been reinforced in their recently published AI delivery plan.

CDA-AMC

Canada’s Drug Agency (CDA-AMC) adopted a closely aligned position in 2025. Building on NICE’s framework, the CDA-AMC highlights the importance of maintaining transparency, trust, and methodological rigor in HTA submissions that include AI-generated evidence. Users remain accountable for all outputs and must justify the use of AI, demonstrate appropriate validation, maintain human oversight, and comply with legal, ethical, privacy, and cybersecurity requirements. The agency explicitly supports a “human-in-the-loop” approach in which AI assists analysts and researchers rather than independently generating evidence for decision-making.

EU HTACG

The EU HTA Coordination Group (HTACG) has taken a similarly pragmatic approach in its 2026 General Principles on the Use of AI in the Preparation of Dossiers for Joint Clinical Assessments (JCAs). The guidance does not restrict the use of AI, including generative AI, but makes clear that Health Technology Developers (HTDs) remain fully accountable for all evidence submitted in a JCA dossier, regardless of how it was produced. The HTACG emphasizes several core principles: accountability, transparency, human oversight, scientific rigor, validation, and traceability. Developers are expected to disclose whether and how AI was used, maintain records of AI-supported workflows, and ensure that all AI-generated outputs are rigorously verified against source data and established methodological standards.

The guidance also stresses that the use of AI does not alter existing evidentiary requirements under the EU HTA Regulation, and that AI-assisted evidence generation must achieve the same standards of completeness, reliability, and reproducibility as conventional approaches. In practice, the position supports the use of AI for activities such as literature review, evidence extraction, evidence synthesis, and dossier drafting, while reinforcing that critical scientific judgement and responsibility must remain with human experts.

The HTACG’s position closely aligns with the broader international consensus emerging from both regulators and HTA agencies: AI can improve efficiency across the evidence-generation process, but trustworthy use depends on robust governance, transparency, validation, and ongoing human oversight.

Regulatory bodies: Enabling AI use, while establishing guardrails

FDA

On the regulators’ side, the FDA has focused its 2025 draft guidance on AI models used to generate evidence that could support regulatory decision-making for drugs and biologics. Rather than evaluating AI technology itself, the agency proposes a risk-based credibility assessment framework centered on the model’s intended context of use. Sponsors are expected to assess and document the credibility of AI-generated outputs through validation, performance testing, risk mitigation, and ongoing monitoring. Importantly, the FDA differentiates between low-risk operational uses of AI and applications that directly influence regulatory evidence, reserving the greatest scrutiny for the latter.

EMA

In Europe, regulators have adopted a similarly cautious but enabling stance. The EMA’s 2024 Reflection Paper explores AI applications across the medicinal product lifecycle, including drug discovery, clinical trials, manufacturing, pharmacovigilance, and regulatory documentation. The paper encourages early dialogue with regulators and emphasizes data quality, model transparency, governance, and lifecycle management as prerequisites for regulatory acceptance.

MHRA

The MHRA has likewise positioned itself as both an enabler and regulator of AI innovation. Its reflection on AI regulation highlights opportunities to improve regulatory efficiency and accelerate patient access to medicines while ensuring that appropriate safeguards are in place. The agency stresses proportional regulation, international harmonization, and evidence-based oversight of AI technologies used in healthcare and medicines development.

A foundation for future AI guidance

Perhaps the strongest signal of international convergence came in January 2026, when the EMA and FDA jointly published ten principles for Good AI Practice in drug development. These principles include human-centric design, risk-based oversight, adherence to regulatory standards, clearly defined context of use, multidisciplinary expertise, robust data governance, sound model development practices, ongoing performance evaluation, lifecycle management, and clear communication. Together, they establish a foundation for future AI-specific guidance and reinforce the view that trustworthy AI requires governance throughout its lifecycle, not simply at deployment.

A global convergence on the responsible use of AI

Taken together, the guidance from NICE, CDA-AMC, the EU HTACG, FDA, MHRA, EMA, and the joint EMA-FDA initiative demonstrates a striking global convergence around the responsible use of AI.

While individual frameworks differ in scope and regulatory context, all emphasize the same foundational principles:

• Human accountability
• Transparency
• Risk-based validation,
• Strong data governance
• Continuous oversight

For HEOR, market access, and HTA professionals, the challenge is no longer whether AI should be used, but how to implement it in a way that maintains scientific credibility, regulatory confidence, and stakeholder trust.

Share Via