The Maryland Public Information Act (PIA) procedures are currently experiencing significant and rapid changes, largely driven by the increasing integration of artificial intelligence (AI) technologies. This technological acceleration brings forth new dynamics, as AI tools not only expedite certain aspects of the process but also introduce challenges stemming from their inherent tendency to misunderstand or misinterpret nuanced components of the legal framework. This article delves into how AI is reshaping the landscape of public information requests, highlighting both the opportunities for efficiency and the critical responsibilities that arise for public officials in navigating these evolving circumstances.
AI's Broadening Impact on PIA Requests
The integration of Artificial Intelligence into the Maryland Public Information Act (PIA) process introduces a new set of dynamics, as highlighted in a presentation at the 2026 MACo Fall Symposium. Alisa Bralove Scherr, Deputy Public Access Ombudsman, detailed how AI is fundamentally altering the way PIA requests are handled, from the initial inquiry stage right through to mediation. A significant challenge arises from AI's capacity to autonomously generate PIA requests. While this automation offers efficiency, it often leads to requests being overly broad, employing vague phrasing such as 'any and all' rather than precise, targeted inquiries. Furthermore, AI systems, by their nature, are devoid of human judgment and perspective, which can prove to be a critical limitation when navigating the complex and often nuanced mediation processes involved in public information requests. This lack of discerning capability underscores a core concern regarding AI’s role as an independent agent in sensitive administrative and legal contexts.
Challenges with Human Requestors Leveraging AI
Beyond AI's direct automation of requests, a different set of challenges emerges when human requestors utilize AI tools to formulate their inquiries or assist in the collaborative production of materials with custodians. A primary concern is the phenomenon of 'AI hallucinations,' where artificial intelligence generates plausible but incorrect information. More profoundly, individuals are increasingly treating AI-generated responses as definitive sources of truth, particularly concerning the Maryland Public Information Act and its intricate legal language. AI models frequently misstate or misunderstand the subtleties of legal terminology, leading requestors to hold confidently incorrect beliefs, as aptly characterized by Deputy Ombudsman Bralove Scherr as bots being 'confidently wrong.' This reliance on flawed AI output complicates communication and resolution between requestors and government agencies, demanding a higher degree of vigilance and verification.
Custodian Concerns and Evolving PIA Records
A significant apprehension voiced by custodians of public information revolves around the inability of current government policy and existing legislation to adapt swiftly enough to the rapid advancements in AI technology. The pace of technological evolution far outstrips the legislative process, creating a regulatory vacuum where guidelines for AI integration are lacking. Furthermore, the very act of using AI by public officials has introduced new categories of accessible PIA records. This includes various forms of AI interactions, such as chat bot exchanges and the data processed by AI algorithms used in public safety applications, like analyzing footage from automatic license plate readers and body-worn cameras. These novel data sources and interaction types were not conceived when the initial PIA legislation was drafted, posing complex questions about data retention, access, and privacy in the digital age.
Best Practices for Navigating AI in PIA
To effectively address the emerging complexities introduced by AI in the PIA process, the presentation outlined several crucial best practices for both requestors and custodians. Firstly, AI should be utilized strictly as a supplementary tool or 'helper,' not as a primary decision-maker. This means that any output generated by AI must be thoroughly verified for accuracy, especially concerning the specific requirements and legal nuances of the Public Information Act. Secondly, there is a strong recommendation to never rely on AI for handling confidential or sensitive information, given its inherent limitations in understanding context, privacy implications, and potential for data breaches. It is also vital to recognize that AI technology, regardless of its sophistication, cannot replace essential human communication or supersede the ultimate authority and discretion of government agencies. Finally, explicitly avoiding the use of AI during the formal PIA mediation process is advised to ensure that critical judgments and sensitive discussions are handled with human oversight and ethical considerations.
