TLDR: The Pennsylvania Supreme Court has adopted an interim policy, effective December 8, 2025, to govern the use of generative artificial intelligence (GenAI) by judicial officers and court personnel across the Unified Judicial System. The policy permits GenAI for tasks like summarizing documents, preliminary legal research, and drafting initial versions, but strictly mandates the use of ‘Secured AI Systems’ for any non-public information to ensure confidentiality and data protection. It emphasizes human accountability for accuracy, ethical compliance, and requires leadership approval for AI tools.
The Supreme Court of Pennsylvania has issued a landmark Interim Policy on the Use of Generative Artificial Intelligence by Judicial Officers and Court Personnel, setting clear boundaries and guidelines for the integration of AI technologies within the state’s Unified Judicial System. The policy, which was issued on September 9, 2025, and will become effective on December 8, 2025, aims to harness the efficiency benefits of GenAI while safeguarding confidentiality, accuracy, and ethical standards across all levels of the judiciary.
Scope and Key Definitions
The policy applies broadly to ‘Personnel of the Unified Judicial System,’ encompassing state-level court employees, judicial officers (including senior judges) and their staff, employees of Supreme Court-established boards, and county-level court employees. It also extends to personal devices used for work-related purposes.
Central to the policy are the definitions of AI systems. ‘Generative Artificial Intelligence’ (GenAI) refers to algorithms and computer processes that generate text, audio, or images based on user prompts. A critical distinction is made between ‘Secured AI Systems’ and ‘Non-Secured AI Systems’. A ‘Secured AI System’ is defined as one that protects the confidentiality and privilege of all user-entered data and documents, does not retain, transfer, or sell such data to third parties, and does not expose it to the public domain. This includes ensuring subcontractors adhere to these requirements. Conversely, a ‘Non-Secured AI System’ is any system that fails to meet these confidentiality and data protection standards.
Permitted Uses and Strict Prohibitions
Under the new guidelines, personnel are authorized to use GenAI for specific work-related tasks, provided the tools are approved by judicial ‘Leadership’ (Chief Justice, President Judges, and Court Administrator). Permitted uses include:
Summarizing documents.
Conducting preliminary legal research, but only with GenAI tools trained on comprehensive, up-to-date collections of reputable legal authorities.
Drafting initial versions of documents such as communications and memoranda.
Editing and assessing the readability of public documents.
Providing interactive chatbots or similar services to the public and self-represented litigants.
A cornerstone of the policy is the strict prohibition against sharing ‘Non-Public Information’ with ‘Non-Secured AI Systems’. Personnel are explicitly instructed to assume that any information entered into a non-secured system will not be treated as confidential or privileged. ‘Non-Public Information’ includes any data restricted by federal or state law, policy, or other legal authority.
Human Accountability and Ethical Obligations
The policy places significant emphasis on human oversight and accountability. Personnel are required to comply with all existing ethical and professional conduct rules, including the Code of Judicial Conduct and Rules of Professional Conduct, when using GenAI. They must become proficient in the technical capabilities and limitations of GenAI tools and maintain this competence.
The policy explicitly warns that GenAI systems ‘may not consider nuances humans take into consideration,’ can be ‘trained on biased material and generate biased content or harmful material,’ and may produce ‘inaccurate because they have little or no basis in fact or reality (commonly referred to as “hallucinations”).’ Therefore, personnel are ultimately responsible for the accuracy, completeness, and ethical implications of any GenAI-assisted work product. Every GenAI output must undergo human review for accuracy, completeness, and potential bias.
Implementation and Oversight
Judicial Leadership is tasked with ensuring compliance, which includes rigorous due diligence in reviewing vendor contracts and end-user licensing agreements for both new and existing GenAI tools. This review must verify that vendors and their subcontractors adhere to the ‘Secured AI System’ requirements, particularly regarding data retention, training on user data, security of transmission pathways, and non-exposure of content to the public domain. Supervisors may also require personnel to seek approval for GenAI use or disclose its use in their work product.
Also Read:
- Irish Courts Affirm AI Use for Legal Correspondence Amidst Broader Judicial Scrutiny
- Maryland Grapples with AI ‘Hallucinations’ in Legal Cases, Spurs Legislative Push for Guardrails
This interim policy represents a proactive step by the Pennsylvania Supreme Court to navigate the complexities of emerging AI technologies, aiming to leverage their benefits while upholding the judiciary’s core principles of justice, confidentiality, and public trust. It sets a precedent for how judicial systems can responsibly integrate AI, with a clear focus on security, ethics, and human oversight.
