From the Public Safety edition of October 25, 2026
UK public dialogue on AI in policing: 98 participants accept AI that supports officers but draw red lines at AI prioritising 999 calls and behaviour monitoring
Home Office, PoliceAI and UKRI (AHRC); delivered by Thinks Insight & Strategy · Policing / public attitudes to AI-assisted decision-making · England and Wales (workshops in Manchester, Norwich and Swansea)
- Publisher
- UK Research and Innovation
- Original publication
- Published by UKRI Oct. 23, 2026 (report dated September 2026)
- Source retrieved
- 2026-10-26
- Event date
- 2026-10-23
What happened
A deliberative dialogue commissioned by the Home Office and PoliceAI with UKRI, and run by Thinks Insight & Strategy, put three AI use cases to 98 members of the public in Manchester, Norwich and Swansea over five weeks in June and July 2026: online crime reporting, 101/999 call handling and CCTV crime detection. Participants were not anti-AI but their support was conditional on clear public benefit, meaningful human involvement and proportionality. They drew clear red lines around AI prioritising 999 calls and AI behaviour monitoring, both hypothetical. Their recommendations: AI should support, not replace, police judgement; people should be told when AI is used and be able to challenge decisions; systems should be tested before deployment and independently evaluated; and an independent body with regulatory powers should audit police AI use. The report notes AI triage of 101 calls became current practice in August 2026, after fieldwork. Facial recognition was out of scope.
Why it matters
One of few structured public-consent studies tied to a national police AI programme; it gives agencies a concrete list of conditions (human contact, notification, challenge routes, independent audit) that communities may expect before AI call handling or CCTV analytics.
Evidence and measured results
Opened the UKRI publication page (published 23 October 2026) and the 1 MB PDF; read the foreword, executive summary, methodology, use-case table, recommendations and conclusions.
Limitations and uncertainty
Qualitative deliberative research with a sample broadly reflective of, not representative of, the public; the acceptability ranking was a temperature check, not a quantitative measure. Commissioned by the bodies promoting police AI adoption. Facial recognition and operationally sensitive uses were excluded. UK legal context limits transfer to U.S. agencies.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-26; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Agencies adopting AI in crime reporting, call handling or CCTV review risk losing public confidence if human contact and accountability are removed.
- Stakeholders
- Chief officers
- Police and crime commissioners or local oversight bodies
- Call-handling managers
- Data protection officers
- Community groups
- Discovery questions
- Which of your AI use cases remove a human from the first point of contact?
- Can people find out AI was used and challenge the result?
- Who independently evaluates the tool before it scales?
- Value hypothesis
- Building the report’s conditions into a deployment could reduce public opposition and give decision-makers a defensible record.
- Potential engagement
- Community-acceptability and safeguards review of one proposed AI use case, ending in a list of required human touchpoints and disclosures.
- Unsupported claims
- Effectiveness of any AI tool
- U.S. public attitudes
- That meeting these conditions guarantees acceptance
Pre-sales engineering
Role takeaway
- Fit
- Police non-emergency contact, online reporting and CCTV analytics programmes.
- Architecture and integration
- Human handoff, AI-use notification and appeal handling must be built into contact centre and reporting platforms.
- Prerequisites
- Defined use case and public benefit
- Human escalation path
- Evaluation plan
- Constraints
- UK legal framework (DPA 2018 Part 3) differs from U.S. law
- Excludes facial recognition
- Qualitative evidence only
- Security
- Limit solely automated decisions; protect data gathered by chatbots and CCTV analytics under data protection rules.
- Proof of value
- Proposed: measure share of AI-handled contacts that reach a human when requested, and time to reach them
- Proposed: test that AI-use notices are presented to every caller or online reporter
Delivery
Role takeaway
- Work
- Add public engagement, notification and challenge processes, and independent evaluation to AI deployments in public contact and CCTV review.
- Dependencies
- National guidance from PoliceAI and the Home Office
- Data protection impact assessments
- Vendor support for human handoff
- Ownership
- Chief officer, with the force data protection officer
- Skills and adoption
- Train call handlers and reviewers to supervise AI output and handle challenges.
- Governance checkpoints
- Public engagement before deployment
- Independent evaluation before scaling
- Entry on the national police AI register
- Acceptance criteria
- Proposed: evaluation published before expansion
- Proposed: every AI-handled contact offers a route to a person
- Risks
- Treating legal compliance as consent
- Scope creep into behaviour monitoring
- Unequal trust across communities
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
The use cases range from AI prompts to human call handlers, to AI as first point of contact for 101, to AI-flagged CCTV review; acceptability fell as human involvement fell.
Governance
Who approves, reviews and stays accountable for outcomes?
Participants want independent evaluation and an independent body with regulatory powers; PoliceAI says it will keep a public register of operational police AI and the checks done before deployment.
Security and privacy
What data, permissions and controls need testing?
Behaviour monitoring and online content scanning raised self-censorship and surveillance concerns; the report notes Part 3 of the Data Protection Act 2018 limits solely automated significant decisions.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Participants expected time saved to go to victims and to call-handler wellbeing, and worried about deskilling and job losses; a route to a human must remain.
Procurement
What should contracts, pricing and exit terms secure?
PoliceAI says it will use the evidence to give chief officers stronger evidence for procurement and deployment decisions.
Operating model
Which teams own the service once it runs?
Operational decisions remain with chief constables; PoliceAI supports national testing, benchmarking and adoption.
Publication history
- 2026-10-25Public Safety · Issue 475 resources
Stable resource ID: uk-ai-in-policing-public-dialogue-2026-final-report-98-participants-20261023