The US government plans to use artificial intelligence to help redact documents connected to FOIA requests, with the Department of Homeland Security moving first on a large scale. The goal is simple: help staff find information that should be removed, save time, and make the review process more accurate.
By the end of September, the DHS plans to use AI on documents that represent over 10 percent of its FOIA requests. That means around 100,000 to 140,000 documents in all, giving the new system a large test across government records.
How DHS plans to use AI for redactions
The DHS will use Google’s AI tools to recommend information that FOIA officers should redact. The system will not make the final decision on its own. DHS employees will review cases after the AI makes its recommendations, checking that the information removed is accurate and complete.
That human review is central to the plan. AI will point staff toward information for possible redaction, while employees will remain responsible for checking the results. The Department of Homeland Security has described the effort as a way to “revolutionize” how the agency handles redactions.
The government says the automation will save time, and the plan description says “this automation is expected to significantly reduce processing times.” Faster processing could help the DHS handle a large share of its FOIA workload with the same review process, but the plan also raises questions about what the systems might remove.
Government agencies are presenting these tools as a way to improve efficiency and accuracy. At the same time, the use of AI could lead to more blacked-out or missing documents if the recommendations remove information that should have remained available.
AI redaction efforts are spreading across agencies
The DHS is not the only agency testing AI for this work. The Interior Department uses Microsoft tools to redact potential attorney-client information, adding another approach to the government’s growing use of automated document review.
The Department of Justice is using a Veriton tool called aiWARE to redact video. That effort shows that the government’s plans extend beyond written records, since video can also contain information that agencies decide to remove.
Health and Human Services is developing its own tool, FRED, to redact unknown information. The FDA is also exploring limited uses of AI to help FOIA staff process records more efficiently. An FDA spokesperson described the effort this way: “FDA is responsibly exploring limited uses of AI to help FOIA staff process records more efficiently.”
These projects do not all use the same tools or handle the same kind of information. DHS plans to use Google’s AI tools for recommendations, Interior uses Microsoft tools for potential attorney-client information, the DoJ uses aiWARE for video, and HHS is building FRED for unknown information.
Speed and secrecy remain competing concerns
The central tradeoff is easy to understand. AI could help employees work through large numbers of records, especially when documents require repeated checks for information that may need redaction. But a system that makes the process faster can also make mistakes faster if employees do not catch them during review.
The DHS plan includes employee review, which is meant to protect the accuracy and completeness of each decision. Still, the scale of the rollout matters: the agency expects to apply AI to around 100,000 to 140,000 documents, covering over 10 percent of its FOIA requests by the end of September.
A FOIA expert at the University of Florida offered a direct warning about the risk: “I think we have to watch the government really carefully on how they use AI to redact records, because I think it’s going to be a very powerful tool for secrecy.”
That concern puts attention on the details of each system’s recommendations and the reviews that follow them. The government wants AI to reduce processing times, while critics worry the technology may create more blacked-out records or missing information. The DHS rollout will show how those goals collide when automated recommendations meet public records decisions.
For now, the government’s approach combines AI recommendations with employee checks. Google, Microsoft, Veriton, and agency-built tools are already part of separate efforts across DHS, the Interior Department, the DoJ, HHS, and the FDA. The result could be a faster FOIA process, but the quality of the records people receive will depend on what the systems flag and how carefully employees review each case.
Based on
