Prince William County, Virginia has implemented the Liminal platform to help employees use generative AI tools without risking exposure of sensitive or personally identifiable information, aiming for compliance across 20 departments.
Prince William County in Virginia has introduced a new platform designed to help government employees use generative artificial intelligence tools while protecting sensitive data. The county selected the Liminal platform after a pilot phase, aiming to ensure that staff can benefit from AI technology without risking the exposure of confidential or personally identifiable information.
The system works by intercepting prompts submitted by employees to generative AI platforms. Before any data leaves the county's network, the platform automatically redacts or removes sensitive information, ensuring compliance with all relevant security standards. Once the AI generates a response, the platform reinserts the sensitive data, allowing employees to receive complete and contextually accurate answers without compromising privacy.
How the Platform Works
According to Prince William County officials, the Liminal platform acts as a filtration layer between staff and AI tools. For example, an employee in the Department of Social Services can use the system to strip out personally identifiable information from a report before requesting a summary from an AI model. The platform identifies and redacts sensitive data, sends the sanitized prompt to the AI, and then restores the original information in the AI's response for the employee's use.
This approach was developed in response to concerns from county staff about the widespread use of large language models and the lack of clear guardrails for protecting sensitive data. Dr. Alicia Hart, deputy county executive for government operations, performance and innovation, explained that the method preserves the intent and context of prompts while ensuring compliance with data protection laws and internal policies.
Pilot Phase and Department Rollout
The county began with a "soft launch" of the platform, involving 150 users across several departments. During this pilot, the Department of Information Technology worked closely with Liminal's leadership to review data flows, test redaction and masking features, and evaluate the platform's performance in different scenarios. After the pilot, the IT department presented its findings to executive management and agency leaders, who approved a broader rollout.
As of now, 20 county departments are using the Liminal platform. The county's Communications and Engagement Office has been sharing information about the tool with employees, and department users have become strong advocates for its adoption. County officials say the goal is to eventually make the platform an enterprise-wide requirement, but they are prioritizing buy-in from staff and stakeholders during the rollout.
Broader Context and Next Steps
The adoption of Liminal comes as Prince William County continues to expand its use of AI in government operations. The county recently reported that its AI-powered PWC311 service, known as Will, handled nearly 90,000 interactions with residents in its first year, including almost 40,000 questions routed to the appropriate departments for assistance. County leaders say they plan to further increase innovation and efficiency through responsible AI adoption.
Prince William County's approach reflects a growing trend among local governments to balance the benefits of AI with the need for strong data safeguards. As highlighted in a related report on the importance of privacy and accountability in government data sharing, experts continue to urge public agencies to prioritize security and compliance as they modernize their digital services.
County officials have not announced a timeline for making the Liminal platform mandatory across all departments, but say they will continue to involve end users and agency leaders in the process. The county declined to provide additional details about the initiative beyond what was shared during recent public events and in official statements.
