Skip to content

Disclosure Overload: AI Reforms Take On 1996 -Era Evidence Rules

Peel Solutions

Introduction

A modernisation exercise thirty years overdue is finally underway. The Home Office has confirmed it will legislate to allow artificial intelligence to help police officers review and summarise digital evidence, accepting a raft of recommendations from Jonathan Fisher KC's Independent Review of Disclosure and Fraud Offences. The rules currently governing how officers manage case evidence date back to 1996 — a time before smartphones, cloud storage or a single one of today's major messaging platforms existed. Investigations have moved on considerably since then; some now generate over 500,000 e-books worth of digital material, and the average fraud case runs to more than 4 million documents. The reforms sit alongside the continued rollout of PoliceAI, the recently launched National Centre for Police AI, and form part of what the government describes as its most significant policing reform programme in almost two centuries. For forces already stretched by casework volume, the direction of travel is clear: AI is moving from pilot project to policy.

What is changing, and why now

The current disclosure regime is built on the Criminal Procedure and Investigations Act 1996, supplemented by the Attorney General's Guidelines on Disclosure. Neither was designed with digital material in mind, and both have strained under the growth of phone downloads, cloud extraction and messaging data that now sit at the centre of most investigations. Jonathan Fisher KC's review, Disclosure in the Digital Age, examined this strain directly and produced 45 recommendations aimed at bringing the regime into line with how evidence is actually generated and held today (Independent Review of Disclosure and Fraud Offences).

The Home Office has now accepted the review's headline recommendation: legislating to permit the use of AI in reviewing and summarising material. Under the current system, officers must manually process and write a summary for every file that could conceivably be relevant to a case — a task that scales badly against six-figure document counts. Under the proposed changes, officers will instead be able to use technology to identify, sort and compile material that today is reviewed line by line.

 

The scale of the disclosure burden

The numbers explain why the government has prioritised this. In 2022/23 alone, the Policing Productivity Review estimated that officers spent around 532,000 hours building case files that the CPS subsequently assessed as requiring no further action — time spent on disclosure work for cases that never proceeded (Policing Productivity Review). That figure sits within a wider pattern: fraud investigations in particular now regularly involve millions of individual documents, a volume no manual review process was ever built to handle. The mismatch between digital case volumes and 1996-era process is, on the government's own account, now a direct driver of delay for both victims and defendants.

Where this sits in the wider reform package

The disclosure reforms are not standalone. They follow the June 2026 launch of PoliceAI, the National Centre for Police AI, backed by £75 million of Home Office funding over three years and hosted by the College of Policing (PoliceAI launch announcement). PoliceAI is expected to free up an estimated 6 million hours of police time per year by 2028 — equivalent to roughly 3,000 additional officers — and forms part of the Police Reform White Paper published in January 2026, which sets out the government's plans for a National Policing Service. Disclosure and evidence summarisation was already flagged as one of PoliceAI's first-year priorities; today's announcement gives that priority a legislative route forward.

 

Implementation signals to watch

Legislating for AI-assisted review

The immediate step is legislation to permit AI use in disclosure work. Forces should expect this to be accompanied by statutory guidance clarifying where automated sorting and summarisation is permitted, and where human review remains mandatory — the disclosure decision itself is expected to stay a matter of professional judgement rather than automated output.

PoliceAI pilots and 2027 scaling

Using Home Office funding, PoliceAI will pilot tools capable of automatically generating summaries of digital material, with a stated ambition to scale successful tools across all 43 forces in 2027. The pace of that scaling, and which forces are selected for early pilots, will be worth tracking closely for anyone planning procurement or training around case file management.

A national governance forum for disclosure technology

The Home Office has also accepted Fisher's recommendation to establish a national governance forum for disclosure technology, bringing together policing, the judiciary, prosecutors and government. Its remit will be oversight of emerging tools and assurance that safeguards keep pace with capability — a recognition that disclosure sits at the heart of the right to a fair trial, and that efficiency gains cannot come at the expense of evidential integrity.

Centralised procurement and the Intensive Disclosure Regime

Alongside the AI recommendations, the Home Office is moving towards centralised procurement of police technology, work already underway through the wider police reform programme and the planned National Police Service. Separately, the Serious Fraud Office has welcomed government plans to explore a new Intensive Disclosure Regime pilot — a signal that fraud cases, with their especially high document volumes, may be an early testing ground for the new approach.

Benefits being sought — and the trade-offs

The case for reform is largely about officer time and case velocity: fewer hours spent manually trawling phone downloads and cloud extractions should mean more time for frontline investigation and victim support, and, in principle, faster progress from charge to trial. Ministers have been explicit that the intention is not to replace investigative judgement with automation, but to remove the administrative bottleneck that currently sits in front of it.

The trade-offs are equally real. Disclosure exists as a safeguard, not an administrative formality, and any perception that AI is compressing or skipping proper review will invite challenge — from the defence, from the judiciary, and potentially in case law. The national governance forum is clearly intended to head off exactly this concern, but its effectiveness will depend on how quickly it can produce usable standards, and how transparently forces communicate what AI has and hasn't done in a given case file.

 

Practical implications for forces

        Policy and case file processes

  • Review local disclosure policies and case file templates now, ahead of legislation, to identify where AI-assisted sorting and summarisation could realistically slot into existing workflows.
  • Build a clear internal line between AI-assisted material review and the disclosure decision itself, so accountability for the latter is never in doubt.

    Evidence review tooling and training

  • Where PoliceAI pilots are announced, assess eligibility early — forces that engage with early pilots are likely to shape practical standards rather than simply inherit them later.
  • Plan training for officers and disclosure officers on how to interrogate and sense-check AI-generated summaries, not just how to operate the tools themselves.

Governance and quality assurance

  • Establish or strengthen local quality assurance processes for AI-assisted case files, so output can be audited and defended if challenged at court.
  • Track outputs from the new national governance forum as standards and guidance are published, and feed local experience back into that process where routes exist to do so.

Procurement and vendor management

  • Monitor the shift towards centralised technology procurement, which is likely to change how individual forces select and contract disclosure tools going forward.
  • Where local contracts for case management or e-discovery tools are due for renewal, factor in the direction of travel before committing to long-term agreements.

Risks, mitigations and metrics

The clearest risk is public and judicial confidence — a sense that AI is being used to cut corners rather than cut administrative burden. Mitigation lies in transparency: clear messaging about what AI tools do (sorting, flagging, summarising) versus what remains a human decision (disclosure itself), backed by the governance forum's oversight role. A second risk is inconsistent adoption across forces, with early pilot sites developing capability that others lack until 2027 scaling catches up; shared standards and shared training materials, distributed centrally, would help close that gap.

Useful metrics to track over the coming year include hours saved on disclosure work per case, time from charge to trial for cases using AI-assisted review, and the volume and nature of disclosure-related challenges or appeals. Comparing these against the 532,000-hour baseline identified in the Policing Productivity Review would give forces, and the public, a concrete way to judge whether the reforms are delivering.

 

Conclusion

Three things are worth watching over the next twelve months. First, the legislation permitting AI use in disclosure work, and the statutory guidance that accompanies it. Second, the pace and location of PoliceAI's pilot programme ahead of 2027 scaling — an early signal of which forces will shape practical standards. Third, the establishment and output of the national governance forum, which will determine how much confidence the wider system places in AI-assisted disclosure. Taken together, these reforms mark one of the more concrete steps yet in translating AI ambition into operational policing practice — and for forces managing growing case file volumes, the practical planning window is now.

 

 

Sources

Peel Solutions