The UK government has accepted all 44 recommendations from the National Commission into the Regulation of AI in Healthcare, but questions remain about who is responsible when medical AI causes harm. Raconteur reports that the response points towards stronger checks after approval, when software changes and real-world performance can expose risks that an initial assessment may miss.
Watch Desk analysis
What happened
The government accepted the commission’s recommendations on 6 October. Raconteur says the response favours more post-approval oversight, while responsibility for failures remains unsettled: NHS Resolution says claims are more likely to be brought against the NHS organisation providing care than against an AI developer. Medical Protection’s Dr Sarah Townley warns clinicians could become a “liability sink” for software failures. Those are attributed views, not a court ruling on who would be liable in a particular case.
The article says the Medicines and Healthcare products Regulatory Agency plans to issue draft guidance by December 2026 on predetermined change control plans. These plans would set limits on software updates and require developers to manage their risks. The regulator has also opened applications for the third phase of its AI Airlock testing programme.
Why it matters
An AI tool does not stop changing, or affecting care, once it clears an approval process. Performance can vary between hospitals, and updates or changes in working practices may alter how a system behaves. Raconteur cites a 2026 study of four deployed clinical AI systems at one healthcare organisation that found performance changed over time, including without deliberate software updates.
For hospitals, the practical advice reported by Raconteur is to keep supplier contracts, records of which software version was used and notes on whether staff followed the AI’s advice. Such records may help explain what happened, but they cannot by themselves settle a dispute over responsibility.
Our read
The useful shift is from asking whether a medical AI passed a check to asking how its performance is monitored once it meets real patients. Better records and change controls could make problems easier to spot and investigate. They do not resolve the harder question of how responsibility should be shared between clinicians, hospitals and developers when no one party controls the whole system. Healthcare AI needs an aftercare plan, not just an approval stamp.
What to watch
- Whether the government publishes an implementation roadmap in spring 2027.
- What limits and safeguards appear in the MHRA’s draft guidance on software changes.
- Whether hospitals adopt version tracking and records of how staff used AI advice.
Discussion spark: When medical AI fails after deployment, should responsibility fall mainly on the hospital using it, the developer controlling the software, or be shared according to who could best manage the risk?
Sources and evidence
- Source update (9 November 2026, 13:25 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.