Public Sector & Government · Issue 03 ·
Public Safety
Lancashire's September 8 deepfake-awareness launch supplies a current harm-response example with no measured impact. Newly archived historical sources show officer review missing report errors and a corrections AI project completing software but not its effectiveness evaluation. Three sources and one cross-source interpretation; no generalized savings or justice-outcome claims. Standalone courts and fresh corrections benefits remain gaps; inaccessible Maryland reporting is excluded.
- Evidence records
- 3
- Cross-source patterns
- 1
- Evidence classes
- 1 independent research1 government audit1 standards or public-body guidance
- Outcomes
- 2 cautionary1 emerging
- Source freshness
- 2 older, newly relevant1 new this fortnight
- Research completed
- 2026-09-09
Choose a role to see its takeaway beside every record in the ledger.
Synthesis · Lighthouse Advisory interpretation
Patterns across the evidence
Evaluation depends on operational preparation as well as working software
The report-editing exercise lacked specific reviewer training, while the corrections project lost its human-subject evaluation after compliance and oversight failures. Prepare the people, approvals and reference data needed for a valid test before treating a functional tool as ready for operational expansion.
Operating questionWhich training, approvals, partner duties and usable evaluation data must exist before this pilot can establish its intended benefit?
Supporting evidenceFederation of American Scientists; Jon PehaU.S. Department of Justice Office of the Inspector General
Full record · every source keeps its link and limitations
Evidence ledger
Police-report research warns that officer review can miss material errors
Peha reports material inaccuracies in generated police reports and missed errors when experienced officers reviewed deliberately flawed reports.
Why it matters, evidence and limitations
- Why it matters
- Historical source newly added to the archive supplies a review-training lens for U.S. police drafting pilots; it is not a new September experiment.
- Evidence and measured results
- The author describes a 2025 CMU exercise using three kinds of generative AI and a separate officer-editing exercise involving hallucinations, omissions and event-order errors. Officers had not received specific AI-editing training. No participant count, numeric error rate or controlled training-effect estimate is supplied.
- Limitations and uncertainty
- This is an organizer's account in a policy memo, not a complete peer-reviewed methods report. A university exercise cannot establish field error rates or prove training fixes the problem. Proposed NIJ programs are recommendations, not verified current services.
Corrections AI audit separates a completed system from an unfinished effectiveness evaluation
OIG found a developed AI intervention system but incomplete deployment and efficacy analysis following human-subject compliance failures and weak oversight.
Why it matters, evidence and limitations
- Why it matters
- Historical evidence newly added to fill the corrections implementation gap: a research partnership with Tippecanoe County demonstrates dependencies beyond software development.
- Evidence and measured results
- The audit reports $1,908,515 spent from a $1,999,778 award. Recruitment reached 61 of 250 planned participants; collected human-subject data were abandoned. OIG used interviews, records and judgmental expenditure sampling, not a causal effectiveness evaluation.
- Limitations and uncertainty
- Nonstatistical audit sampling cannot support population-wide projections. OIG explicitly did not assess application effectiveness. Purdue disputed several recommendations and emphasized technical delivery; OJP agreed with the recommendations. No claim that the tool caused absconding or changed recidivism is justified.
Lancashire launches deepfake-awareness campaign with reporting guidance
Lancashire announced a deepfake-awareness campaign for young people and families, linking prevention advice, reporting and victim support.
Why it matters, evidence and limitations
- Why it matters
- Same-day international example of police response to AI-enabled harm. Included solely in public-safety for policing and victim support; no K12 cross-tag or inference of classroom deployment.
- Evidence and measured results
- The announcement links a guidance page encouraging trusted-source checks and reporting rather than sharing harmful material. It describes adoption after Essex's launch but supplies no evaluated reach, behavior-change measure or victim-outcome baseline.
- Limitations and uncertainty
- Campaign launch and operator descriptions are not evidence of reduced abuse or reliable public deepfake detection. UK reporting and legal language must be adapted by qualified local teams; no U.S. legal obligation is inferred.
How to read this edition
Source findings, measured results and limitations come from the cited publications. Patterns, operating questions, role takeaways and implementation considerations are Lighthouse Advisory interpretation, stated as questions to validate locally rather than guaranteed outcomes. Vendor and operator claims are labeled as claims. Full research method.
- Independent research
- Research conducted outside the implementing organization.
- Government audit
- An oversight review of performance, controls, or operations.
- Standards or public-body guidance
- Normative or advisory guidance from a standards body or public institution.