The healthcare product sector is experiencing a technological transformation like never before. Algorithms analysing radiology images, natural language models summarising complex clinical records—the use of artificial intelligence (AI) is no longer a future promise but a central driver of operational efficiency.
Yet the introduction of Regulation (EU) 2024/1689, known as the AI Act, brings a regulatory challenge that many companies have yet to fully integrate into their processes.
Dual marking and risk classification
A key challenge is the so-called “dual marking” issue. Most AI systems used in healthcare, particularly those affecting patient safety or forming part of healthcare products, are automatically classified as High-Risk AI Systems under the AI Act.
This means compliance goes beyond standard MDR or IVDR obligations: companies must also demonstrate algorithmic robustness, cybersecurity, and transparency. The lack of initial alignment between notified bodies and AI supervisory authorities creates uncertainty about whether existing technical documentation will be accepted in future assessments.
The challenge is not just technological—it’s proving that your systems meet all standards without slowing development.
The consequences of inaction
Ignoring these requirements is not an option. Non-compliance can carry serious consequences:
- Massive fines: up to €35 million or 7% of global turnover.
- Loss of CE marking: without it, the product cannot be marketed in the EU.
- Reputational and legal risks: a biased algorithm may be treated as a systemic oversight failure, exposing companies to liability.
- Launch delays: establishing a quality management system integrating AI with GxP standards takes considerable time and resources.
Every day without a clear plan increases operational and commercial risk.
How to prepare: Compliance by Design
The most effective strategy is not to wait for final technical guidance. Leading companies embed the AI Act into product development from the outset.
Key steps include:
- Comprehensive AI system audit: identify all AI systems in use or development and distinguish between high-risk applications and limited-risk tools.
- Integration with quality management systems (ISO 13485 / GxP): incorporate data governance, live technical documentation, and human oversight into a single workflow.
- Continuous monitoring: track AI performance and model drift to ensure ongoing safety and compliance.
Anticipation is more valuable than reaction.
The value of a specialised partner
Managing translation, localisation, and technical documentation for complex AI systems is not trivial. A specialist partner can deliver value in multiple ways:
- Ensuring technical information, warnings, and usage guidelines are correctly adapted for every language and market.
- Preparing documents ready for audit without requiring reprints or last-minute corrections.
- Integrating final artwork, layout, and symbols even in non-Latin scripts, preventing critical design errors.
In short, it is not just about translating text—it is about delivering final documentation that is production-ready and audit-compliant, reducing risk and accelerating time-to-market.
Global guidance and best practices
For a broader perspective, companies can rely on the European Commission AI Guidelines and international frameworks such as the WHO’s principles for AI in health. These offer globally recognised recommendations on algorithm transparency, risk assessment, and data governance, applicable across multiple jurisdictions.
Embedding these principles from the start reduces rework and ensures AI systems comply with both European and global expectations.
Post-market surveillance and adaptation
Approval is not the end. AI systems can experience model drift—changes in clinical protocols or patient populations can affect outputs. A robust post-market monitoring plan allows companies to detect deviations early, linking AI telemetry to risk management plans and ensuring patient safety.
Patient safety and regulatory compliance are ongoing processes, not one-off checkpoints.
Conclusion
The AI Act should not be seen as a bureaucratic hurdle. For companies that understand it, it is an opportunity to demonstrate leadership in innovation, ethics, and safety. Those who integrate regulatory requirements from design stage and collaborate with partners capable of handling translation, localisation, and multilingual production-ready documentation will reach the market faster, with fewer risks, and gain the confidence of auditors and users alike.
In the new era of digital healthcare, compliance is the currency of sustainable innovation.
