Recent Right to Know Law Rulings: Defining "Agency Records" in the Age of AI
The Pennsylvania Office of Open Records (OOR) is starting to issue decisions concerning the Right to Know Law (RTKL) and Artificial Intelligence (AI) which we wish to bring to your attention.
Hamill v. Honesdale Borough
The first case is Hamill v. Honesdale Borough. In this matter, James Hamill submitted a RTK request to Honesdale Borough, seeking “the full, unedited recording of the PUBLIC SAFETY /ZONING & CODE ENFORCEMENT meeting from Tuesday, May 13 2025 @ 03:15PM EDT as recorded by read.ai. A report was generated and available as of June 17th to the Borough Secretary.”
The borough denied the request and Hamill appealed the denial to OOR. The borough argued that no authorization or permission was given by the borough for Read.ai to record the meeting or to automatically send a “Pre-Read” summary to the borough council members. The borough also argued that the Read.ai recording of the May 13, 2025 Public Safety/Zoning & Code Enforcement Committee meeting was not a record of the borough because it was not created by the borough, but was created by software without the borough’s knowledge or consent and would not be used for any business of the borough.
In support of its position, the borough addressed whether:
· The relevant committee meeting was open to the public (it was not)
· The borough was in possession of the requested recording generated by Read.ai (it was not)
· The borough subscribed to or otherwise utilized the software suite that generated the recording and transcript summary referenced in the request (it did not)
According to the OOR, Read.ai is an AI-powered assistant and productivity platform that automates meeting notes, transcribes conversations, and generates summaries of meetings with action items, highlights, and key questions. It accomplishes this, in part, by attending meetings as an independent attendee - recording, transcribing, and generating insights of the contents of the meetings it attends. According to its website, Read.ai will auto-join all meetings if the auto-join setting is enabled, and the auto-join setting is a default one. When connected to a calendar, Read.ai will automatically join meetings on that calendar and will also automatically send out transcript summaries of meetings based on the audio recording captured by Read.ai when it attends a meeting as a participant.
The issue in this appeal was whether the audio or video recording of a non-public agency meeting by an AI productivity platform on the agency’s computers—but without the knowledge or consent of agency personnel—meets the definition of a record of the agency. The RTKL uses a two-part inquiry for determining if certain material is a record:
1) Does the material document a “transaction or activity of an agency?” and
2) If so, was the material “created, received or retained in connection with agency business.”
The OOR determined that the recording of the May 13, 2025 committee meeting documented an activity of the borough. The question, then, was whether the recording was created or received by the borough in connection with a transaction, business, or activity of the borough. The location of the record or even an agency’s possession does not guarantee that a record is accessible to the public, but rather the character of the record controls.
The borough argued that the participation of Read.ai in the May 13, 2025 committee meeting and the subsequent emailing of the transcript summary to borough council members was the result of the settings of the Read.ai platform and not the intentional actions of borough personnel.
The recording of the May 13, 2025 committee meeting was the result of software “attending” and recording the meeting as an independent participant without the authorization of the borough. The borough stated that it was not in possession of the recording generated by Read.ai and that it had never contracted with or authorized Read.ai to perform meeting transcription or recording functions for the borough.
Based on the evidence in this appeal, the OOR concluded that the Read.ai recording of the May 13, 2025 committee meeting was akin to an unauthorized recording in that, even though it was a recording of an agency meeting:
1) The recording was not created or authorized to be created by a borough official acting in their official capacity, but instead the borough was unaware of its existence until the “Pre-Read” email was sent out by Read.ai, and
2) The borough had not utilized the recording for any official business of the borough or for any official purpose, nor had the Read.ai recording and transcript been created or adopted by the borough for any official business.
Accordingly, the Read.ai recording did not meet the definition of a record of the agency subject to access under the RTKL.
Nuckols v. Allegheny County
The second case is Nuckols v. Allegheny County. In this matter, the requester sought all chat histories, conversation logs, and usage records from a number of AI platforms, including but not limited to ChatGPT and Microsoft Copilot used on county owned devices or used by county employees for county business.
OOR stated that because of the “rapidly evolving nature of generative AI” it would “adjudicate each appeal on a case-by-case basis, guided by the unique facts and evidentiary record of each matter.” Based on the evidence, OOR concluded that the AI records were created for personal use regarding county-related work, but the output was not used for official county purposes, nor shared with other employees. Therefore, the OOR determined that the records did not have to be released because they were the functional equivalent of working papers and personal notes.
Both of these cases point out the importance of knowing how AI is being utilized as the facts of each situation will determine whether records generated by an AI platform will be determined to be a public record or not. Also, as AI is used more often we predict that RTK requests for records generated by AI will increase.
Clients who have questions regarding issues discussed in this article, or any employment law matter, should feel free to call us at 215-345-9111.