AI is changing both sides of FOIA. Here’s how agencies should prepare.

FOIA teams are under growing pressure as workloads expand and many programs face staffing constraints, technology changes and other resource limitations.

In fiscal year 2025, federal agencies received about 1.7 million FOIA requests, marking a 13% increase from FY2024 and the highest number on record. The Department of Justice alone received more than 27,000 more requests than in FY24, bringing its total to 159,743, while its backlog jumped from 21,567 to 29,309. 

As requests pile up, FOIA professionals struggle to make the determinations generally required within FOIA’s 20-working-day statutory period. One analysis found that, on average, requesters with pending FOIA submissions waited more than 240 days to hear back. Delayed determinations can also increase litigation risk, adding cost and work for already overburdened teams.

AI can help workers parse records, route requests, review documents, support redaction and surface the right information. 

But it’s a catch-22: That same technology can also help requesters research agencies, draft requests and submit them at greater scale. One federal practitioner recently reported seeing 50-100 requests arrive within moments from the same requester, and other practitioners have described similar experiences. These accounts are anecdotal, but they illustrate how AI-assisted submissions could place additional pressure on workflows designed to process requests individually.

What should agencies do to prepare for AI’s influence? 

Three priorities for modern FOIA programs

1. Use AI only when you can verify its work

Where can AI make FOIA processing faster, without making the results more difficult to trust? We recently interviewed dozens of senior practitioners across corporate legal departments, federal agencies and law firms about how AI is being used in legal, FOIA, records, IT and compliance work today. Across our research, the strongest AI use cases shared several characteristics: bounded tasks, known source material, verifiable outputs, risks that can be appropriately managed and clear human accountability. 

For FOIA requests, that points to tasks like:

• Request routing 
• Document summarization 
• Classification 
• First-pass analysis 
• Identification of potentially responsive material for further review 

The closer AI gets to consequential decisions, the more human oversight the task should demand. Faster FOIA processing should never come at the expense of accuracy, consistency, or accountability. Final redaction and withholding decisions, for instance, are higher-risk applications and should remain subject to review and approval by qualified FOIA professionals.

With appropriately governed tools and processes, agencies can strengthen their ability to understand and document how records were identified, reviewed and ultimately released or withheld.

2. Prepare for AI-assisted volume by expanding access proactively

Historically, many FOIA workflows were designed around an individualized process: A requester seeks specific information, and the agency receives, assesses, routes and responds to the request. AI can reduce the effort needed to research agencies, formulate requests and prepare submissions. As a result, processes designed to handle requests strictly one at a time may need to evolve if submission volumes increase significantly.

Agencies should assess whether their existing tools and processes can manage higher submission volumes and identify where targeted changes are needed. Asking FOIA professionals simply to work faster is unlikely to resolve persistent capacity constraints. Instead, agencies should prepare strategically for AI-assisted volume through tactics such as:

• Strengthening intake and triage 
• Offering clear requester guidance 
• Expanding proactive disclosure 
• Where legally permissible and operationally appropriate, grouping similar or recurring requests for coordinated processing while preserving each requester’s applicable statutory rights and deadlines.
• Making existing public information easier to find

3. Design AI with federal constraints in mind 

Federal AI modernization can’t start with the tool alone. Federal agencies operate with legal, records management, privacy, security, access-control and transparency requirements that differ materially from private-sector environments. Many of those constraints influence what an AI system can access, where it can operate, what it can create and what ultimately happens to its output. AI deployments should account for these requirements from the outset. 

Federal teams may not be able to adopt private-sector AI tools without modification or additional controls. Agencies should design pilots and deployments around the legal, security, privacy, records management and operational constraints that determine whether a tool can be used responsibly.

For example, agencies should consider how AI-generated summaries and other outputs are used and maintained, whether they constitute federal records and what records management and disclosure requirements may apply. They should also establish governance before deployment around what AI-generated materials are created, how they are used and retained and how they are associated with source records.

Additional considerations should include:

• Authorization level 
• Data sensitivity 
• Records treatment
• FOIA impact 
• Clear audit trails 
• Controlled access 
• Human review

FOIA modernization has to address both sides

AI tools are equipping agencies with meaningful new capabilities for managing rising FOIA workloads. Yet technology alone cannot solve the capacity challenge, particularly if easier request preparation and submission contribute to additional volume.

This next phase of FOIA modernization requires agencies to think beyond processing requests faster. Agencies need to use AI where its work can be verified, establish clear governance and expand proactive access to information. The goal is to reduce the burden on strained teams while keeping FOIA professionals firmly in the driver’s seat and making public information easier to access before another request enters the queue.

Varun Bisht is the vice president of governance and compliance for Casepoint — a leading provider of secure, unified eDiscovery, legal hold, FOIA and investigations solutions.