This month’s position
Capability is moving quickly. Organisational permission and evidence still need to be specific.
Begin with defined, lower-risk professional tasks and use current governance routes. Do not let a new feature quietly widen what staff upload, what the system can access or what the output influences.
Policy & governance
NHS England’s current AI information-governance area brings together guidance for service users, health and care professionals and IG professionals. Services should use it as a route into local assurance, not as approval for a product.
NHS England AI guidance →New research
A 2026 systematic review maps early AI-based approaches to adolescent substance-use prevention. The field remains developmental: external validation, safeguarding, bias and real-world effectiveness need careful attention.
Read the systematic review →AI tool development
AI assistants increasingly work across text, images, audio and connected documents. Broader capability increases the need to understand account type, permissions, data flow and what a connected tool can retrieve.
WHO governance guidance →Practice application
A useful low-risk starting workflow is to turn current public guidance into a staff briefing, then check every action, exception, source and date before circulation.
See the worked example →Watchpoint
Do not assume removing a name makes a case anonymous. A combination of age, location, dates and rare circumstances may still identify somebody.
Read the confidentiality guide →One useful workflow
Create a source-bounded comparison: select two current public documents, define the question, request page references, examine differences and keep the final interpretation human-led.
Open the use case →What to read
UK Government AI Playbook ↗ICO artificial intelligence guidance ↗NICE Evidence Standards Framework ↗Published . Reviewed .