AI disclosure in legislative remarks

Summary#

A-11684, the “compulsory artificial narrative disclosure for official rhetoric (CANDOR) act,” would amend New York’s Legislative Law by adding a disclosure rule for state legislators. Its stated goal is to require disclosure when a member’s remarks were drafted in whole or in part by artificial intelligence (AI). The bill is introduced and is not law.

  • A state senator or assembly member who will speak, read a statement, or enter comments into the official legislative record must first give a verbal disclosure if AI drafted any part of the remarks.
  • The disclosure must be clear and prominent and follow substantially a format that states an approximate percentage of AI-generated text. The proposed wording also says the remarks are not the member’s own “organic intellect, independent thought, or cognitive labor.”
  • If a member does not disclose, the bill requires an immediate point of order and a permanent digital warning on the remarks in the official record.
  • The bill does not prohibit legislators from using AI.

What it means for you#

  • State legislators: They would have to give the disclosure before remarks drafted in whole or in part by AI. The bill does not set a method for calculating the approximate percentage.
  • The public: The official record would carry a permanent digital warning if a member failed to disclose. The bill does not create a new public service or give the public a right to review or challenge a disclosure.
  • Current law: The supplied materials do not include the existing Legislative Law, so I cannot verify what disclosure or record-marking rules already apply.

Money#

No publicly available information on the bill’s cost.

  • The bill requires a permanent digital watermark on non-compliant remarks but does not say who must add it, what system would do so, or what it would cost.
  • It does not set out funding, staffing, fees, fines, or a spending amount.

What is unclear#

  • The bill does not explain how a member should calculate the approximate share of AI-generated text.
  • It does not define when AI helped “draft” remarks—for example, whether editing or suggesting wording counts.
  • It does not say who determines whether a member failed to disclose, or how the required point of order and watermark would be handled.
  • It does not explain how the warning’s claim of a “High Probability of Automated Plagiarism” would be assessed. The bill does not establish that AI use is plagiarism.
  • The bill says it would take effect immediately if enacted, but passage is uncertain.

Case for#

  • The bill appears intended to tell the public when AI helped draft remarks delivered in the Legislature.
  • A possible argument for the bill is that a disclosure in the speech and official record could make AI use more visible to people following legislative debate.
  • Requiring the disclosure before the remarks could make the information available at the time they are delivered, rather than later.

Case against#

  • One concern is that the required wording characterizes AI-assisted remarks as not the member’s own independent thought or cognitive labor, even when the bill does not distinguish different kinds or amounts of AI assistance.
  • The bill does not provide a reliable way to calculate the required percentage, which could make disclosures inconsistent.
  • The permanent warning uses “plagiarism,” but the bill does not define plagiarism or require a finding that plagiarism occurred.
  • The bill creates consequences for non-disclosure without explaining who decides that a violation happened or how the record marking is carried out.