MHRA seeks updated NHS rules for AI
UK needs new laws for AI in healthcare, says watchdog
Thursday, 10 September 2026
Warm-up
- Where should AI be used in healthcare, and where not?
- What information should patients get before AI is used in their care?
- When have you trusted tech more than a human, or the opposite? Why?
Vocabulary
- watchdog
- An official body that monitors organisations and can demand changes.
- recommendation
- Formal advice suggesting what should be done next.
- regulate
- To control something by setting and enforcing rules.
- commission
- A group asked to study an issue and report findings.
- authorisation
- Official approval that allows something to be used.
- scribe
- A tool or person that records what is said in a consultation.
- bias
- An unfair tilt in data or decisions that harms accuracy.
- penalise
- To punish a person or company for breaking rules.
Reading
The UK's medicines watchdog, the MHRA, says new laws are needed to regulate AI in the NHS and other care settings. It issued 44 recommendations after an independent commission heard from over 12,000 people, including patients and clinicians. The agency expects AI to be used routinely in NHS services, but says trust must be maintained.
The proposals call for continuous monitoring of approved AI tools, and power to withdraw authorisation if performance drifts. Patients should have a clear right to know when AI is used and easy access to information about the product. The MHRA also wants power to penalise developers that miss required standards, and an AI "L-plate" scheme so clinicians can trial new models under close supervision.
Current rules were built for fixed devices such as hip or knee replacements. By contrast, many AI systems learn from new data and can change after approval, which poses new risks.
In clinics, AI "scribes" reportedly help about 40% of UK GPs by recording consultations and drafting reports. Experts warn about bias in training data and unsafe chatbot advice, although others call it an exceptional opportunity for healthcare and research.
Comprehension
- Which UK body is calling for new AI laws?
- How many recommendations did the review produce?
- What right would patients gain under the proposals?
- Why are current rules seen as outdated for AI?
- What share of UK GPs reportedly use AI "scribes"?
Grammar focus
Adverbs and adverbial phrases: degree
Adverbs of degree show how strong or intense something is (e.g. much, very, completely, far). They usually come before adjectives, other adverbs, or comparatives: much more confident, completely different. (Examples constructed.)
- AI performance can **completely** change after approval.
- Clinicians reported they felt **far more confident** during trials.
Grammar exercise
Circle the correct option in each sentence.
- Clinicians felt (much / very) more confident.
- AI tools require (highly / very) careful monitoring.
- Systems must be (completely / far) transparent.
- Some models are (far / totally) more adaptive.
- Developers need (much / completely) stronger safeguards.
Discussion
- Should developers face penalties when medical AI underperforms? Why?
- When should patients be allowed to opt out of AI tools?
- Does AI note‑taking improve care quality, or only efficiency?
- Which safeguard matters most: transparency, oversight, or human control?
Vocabulary, reading, and exercises with instant feedback