Are ChatGPT Records Subject to FOIA?
As federal agencies increasingly use ChatGPT and other generative artificial intelligence tools, a new transparency question is emerging: Are government employees’ AI prompts, responses, and conversation histories subject to the Freedom of Information Act (“FOIA”)? There is not yet a reported federal appellate decision squarely…
As federal agencies increasingly use ChatGPT and other generative artificial intelligence tools, a new transparency question is emerging: Are government employees’ AI prompts, responses, and conversation histories subject to the Freedom of Information Act (“FOIA”)? There is not yet a reported federal appellate decision.
Existing precedent, however, provides a workable framework. In many circumstances, AI records are likely to qualify as “agency records”—particularly when an agency creates or obtains the information and exercises control over it.
What Happened
ChatGPT conversations may present a different situation. An employee might deliberately ask an AI system to analyze a regulation, summarize comments, prepare a draft, or compare policy alternatives and then use the response in official work.
Nate Sanford, Washington City Officials Are Using ChatGPT for Government Work, KNKX Pub.
Complaint ¶¶ 18, 32, Democracy Forward Found.
As generative AI becomes more integrated into agency work, traditional FOIA principles are likely to determine which AI records must be searched, reviewed, and potentially disclosed.
Key Details
Those facts could support agency-record status under the traditional control analysis. The strongest case involves records maintained in an agency-managed or enterprise AI environment where the agency has access to, uses, or controls the information.
It is whether the government created or obtained the information and exercised sufficient control over it.
The central question is therefore not whether AI created the record.
FOIA exemptions may protect predecisional deliberations, attorney-client communications, attorney work product, personal information, confidential commercial information, classified material, or law-enforcement records.
Why It Matters
If an agency maintains usage logs, sets retention rules, or can retrieve or export conversations, those circumstances weigh in favor of agency control. The argument becomes even stronger once an employee saves an AI output to an agency drive, copies it into.
Even if a ChatGPT conversation qualifies as an agency record, disclosure is not automatic.
State public-records developments also provide guidance.
What Reports Say
Coverage of the story so far points to:
Continued reporting by The National Law Review as more details emerge
