On October 19, the Legal Services Corporation published a Federal Register notice revising the criteria it uses to evaluate legal-aid grantees. At the end, the notice included a direct disclosure: “ChatGPT was used to draft this document.” It added that the generated material provided structural and content suggestions, which the author reviewed, adapted, and revised for LSC's needs and tone.
LSC is a federally established nonprofit corporation, rather than a federal agency. The notice was signed by Stefanie Davis, its deputy general counsel. Those details make the disclosure more useful than a bare “AI-assisted” label. Readers can see what the tool contributed, what the human did next, and which institution issued the final document.
The disclosure also sets a boundary. It does not say that ChatGPT chose the policy, evaluated legal-aid programs, or made decisions about services. The revised criteria themselves cover how grantees assess community needs, plan their work, deliver services, and examine the results. They take effect January 1, 2027.
Read the handoff
This is a useful minimum, not a complete disclosure standard. The notice does not identify the ChatGPT version, prompts, machine-generated passages, or edit history. A reader could not reconstruct how a disputed sentence changed from draft to publication.
Still, it records the handoff that matters most: a tool offered material, a human reviewed and changed it, and a named institution stands behind the result.
That distinction will matter as similar labels appear on reports, school materials, workplace policies, benefit notices, and other documents people may need to trust or act on. A bare statement that AI was involved does not locate responsibility. The disclosure should make that responsibility visible.
When you see an AI-assistance note, ask two questions:
- What did the system actually do? Look for a specific role such as drafting, summarizing, structuring, or translating.
- Who reviewed the result and approved the final words? Look for a person, title, office, or institution that remains answerable for the document.
If the label answers only the first question, the accountability trail is still incomplete.