Regulatory
Industry readiness remains uneven.
manufacturers invest heavily in AI governance structures, life-cycle monitoring capabilities, and dataset governance processes,” Mante says. “AI-specific quality system controls and model change management procedures are not optional requirements for OEMs. “As expected, industry readiness remains uneven. Particularly challenging is that many existing quality systems were designed for traditional software/ hardware, [and struggle] to address model drift, changing datasets, explainability, bias monitoring and adaptive AI systems generally. The EU AI Act transforms many previously voluntary practices into legal obligations for high-risk AI systems. “For multinational OEMs operating within the EU and UK, this means compliance programmes are increasingly designed around EU future expectations even where products are marketed in the UK. “Crucially, many organisations still produce compliance evidence retrospectively, yet future regulations increasingly expect governance embedded within design and development from the outset. Similarly, Brexit implies different compliance requirements, like conformity assessments, apply to the UK and EU.”
420
In a head-to-head comparison of AI-controlled vs human-controlled versions of the same fluoroscopic navigation platform, accuracy was similar for cup position and leg- length discrepancy.
Arthroplasty Today 14
For high-risk systems, the AI Act requires a documented quality management system and covers design control, validation, data management, risk management, record-keeping and post-market monitoring. Existing sectoral quality systems can incorporate these elements rather than necessarily operating as a separate structure.
The UK framework is also still moving. The MHRA’s Software Group works on software and AI as a medical device, while the government is introducing wider medical-device reforms through secondary legislation. NICE’s Evidence Standards Framework can support evidence generation for digital health technologies, but it is not itself a regulatory approval route and its standards are non-mandatory.
The UK debate is evolving. In June, the MHRA published evidence gathered for the National Commission into the Regulation of AI in Healthcare,
drawing on patients, healthcare professionals, industry, academics and other health-system stakeholders. Its work is intended to inform future rules, underlining that the domestic framework remains in development rather than settled. For OEMs, the direction is nevertheless towards stronger life-cycle evidence: data governance, human oversight, change control and post-market monitoring need to be planned earlier in development rather than assembled at the end.
Looking, legally
Dr Colm McGrath, reader in comparative law at King’s College London, sees the same challenge through the interaction between UK and EU regimes. The UK’s post-Brexit position creates room for regulatory divergence, but companies selling into the EU remain exposed to EU requirements. The updated EU Product Liability Directive also expressly treats software, including AI systems, as products for liability purposes. McGrath sees the competing policy ambitions clearly: “It may be that as corollary risks of widespread AI use in the health sector are clarified, a more sober note will have to be struck and compromises made between lofty national aspirations and more personal matters of individual and collective patient tolerance and expectation.”
The question is not only whether AI-enabled devices work, but what evidence should be required about how they work, how decisions are reached and what happens when models, data or clinical contexts change. “We need to find a way to fairly balance all the relevant interests here and, currently, that’s far from obvious/straightforward,” McGrath concludes. “But it’s fair to say we can at least see consensus on some necessary foundational values beginning to emerge, like transparency.”
For OEMs, that consensus is becoming operational. The compliance cliff is not a single deadline. It is the point at which claims about responsible AI must increasingly be backed by systems, records and evidence that can withstand scrutiny throughout the product life cycle. ●
www.medicaldevice-developments.com
Generative AI/Adobe Firefly
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88 |
Page 89 |
Page 90 |
Page 91 |
Page 92 |
Page 93 |
Page 94 |
Page 95 |
Page 96 |
Page 97 |
Page 98 |
Page 99 |
Page 100 |
Page 101 |
Page 102
