Provenance & Intent
A submission has always been read as the words of the person who sent it. Now a user can file through an agent that drafts for them, and an agency can read through an agent of its own. Nothing in the text shows how a submission was produced, or whether anyone holds the view it contains.
So two things have to be established from outside the text. The first is origin: where the submission came from and under whose authority. The second is intent: whether the words are a position the sender holds, or one an agent generated and forwarded in their name. Both have to be recorded when the submission arrives, because neither can be recovered from the text afterward. Where the agency reads through an agent of its own, it takes on a second record to keep: what that agent did with the submission it was given.
Policy challenge
A government channel that accepts open digital input must keep acting on what it receives, yet cannot authenticate where any submission came from, by what means, or on whose behalf.
As agent-mediated drafting and filing become ordinary, unverifiable origin shifts from an edge case to the channel's default condition. Detecting machine authorship after the fact is a losing arms race, which leaves the agency holding a duty to consider submissions it has no dependable way to trace to a person or an intent.
Design challenge
Capture signals about who produced a submission, how, on whose behalf, and under what authority.
Distinguish a personally held position from a forwarded one.
Make capture effective anywhere along a spectrum from pure self-attestation ("I declare") to cryptographic verification ("I can prove"), without inspecting content.
Keep the path open for anyone who can't produce the stronger signal, so capture never becomes the exclusion mechanism.