#49 - AI in Law Enforcement / Evidence & Governance / PoliceReports.ai

Link to Folder with Everything

2026.07.29.officehours.49 — Google Drive folder

Recording of the Video

Summary (by Gemini)

Summary

This meeting explored Artificial Intelligence in law enforcement governance, legal evidentiary standards, and sustainable business model strategies.

Law Enforcement AI Integration

Industry experts emphasized prioritizing officer-controlled dictation over automated capture to ensure narrative integrity. Effective deployment requires deep understanding of specific agency workflows rather than generic solutions.

Legal Frameworks and Compliance

Current discussions identified significant gaps in federal oversight regarding Artificial Intelligence admissibility in court. Practitioners underscored the necessity of strict adherence to Criminal Justice Information Services compliance standards.

Sustainable Business Growth Strategy

The firm adopted an end-user centric development roadmap while maintaining a bootstrapped model to ensure operational independence. Leaders prioritized practical workflow optimization over buzzword-driven Artificial Intelligence deployment.

Next steps

• [Chris Ryan, Lili Kazemi] Create thought leadership: Draft follow-up content on AI governance and bespoke investigative strategies.
• [Rahul Singh] Share grant information: Forward potential research grant opportunities concerning AI safety and governance to the speakers.
• [Rahul Singh] Upload meeting video: Publish the recording of this session to the organization YouTube channel.

Details

• Meeting Logistics and Setup: Rahul Singh, Lili Kazemi, and Chris Ryan initiated the call by coordinating access for attendees, verifying screen-sharing capabilities, and establishing communication preferences for the audience (00:00:03). They agreed that while the chat would remain open, they encouraged verbal interaction or raised hands to facilitate a dialogue during the presentation (00:01:58).
• Intelcraft Community Introduction: Rahul Singh opened the session by introducing Intelcraft, a community of AI practitioners focused on practical, non-hyped applications of AI, such as AI sovereignty, security, and safety. Rahul Singh highlighted that previous sessions covered AI in real estate and finance, and noted that the community utilizes a WhatsApp group for discussions and a calendar for event tracking (00:04:38).
• Core Meeting Objective: Lili Kazemi introduced the session's goal: exploring the intersection of AI and law enforcement, particularly regarding the evidentiary and legal responsibilities when AI becomes involved in police work (00:07:05). The discussion focused on human-in-the-loop versus human-on-the-loop interactions and how to effectively account for responsibility in criminal justice (00:08:31).
• Speaker Background and Company Origin: Chris Ryan, Chief Product Officer at PoliceReports.ai, provided their background in law enforcement, including serving as a patrol officer, narcotics detective, and patrol commander (00:09:40). Chris Ryan explained that PoliceReports.ai was founded by former law enforcement officers to address the arduous paperwork burden of police work by leveraging AI, focusing on efficiency and proper governance (00:10:40).
• Law Enforcement Workflow Variability: Lili Kazemi and Chris Ryan discussed the vast differences between law enforcement agencies across the U.S., noting there are approximately 18,000 agencies that operate with diverse workflows and records management systems (00:13:01) (00:15:36). Chris Ryan emphasized that for AI to be effective, agencies must deeply understand their specific workflows rather than treating AI as a "one size fits all" solution (00:16:59).
• Historical Context of Police Reporting: Chris Ryan reviewed the evolution of police paperwork, from typewriters and paper forms in the 1980s and 1990s to modern records management systems (00:20:39). Despite technological advancements, the industry remains heavily burdened by documentation tasks, often requiring officers to write reports in their cars during downtime (00:22:11).
• AI Implementation Strategies: Chris Ryan distinguished between two approaches to AI document completion in law enforcement (00:23:20). The first, which Chris Ryan discourages, relies on body-worn camera audio/video; this is problematic because it captures extraneous background noise and conversation, potentially introducing liability and evidentiary issues regarding hearsay and the officer's cognition. The second, which PoliceReports.ai utilizes, involves officer-controlled dictation that allows the officer to articulate observations, thoughts, and feelings, thereby maintaining the integrity of the narrative while gaining efficiency (00:24:44) (00:30:31).
• Evidentiary Concerns and Witness Authentication: Lili Kazemi and Chris Ryan discussed the challenge of authenticating witness interviews when using AI (00:31:36). They noted that rather than relying solely on AI-generated synopses, modern practice favors retaining original recordings—such as audio or body-worn camera footage—as the primary source of evidence, ensuring that AI-generated reports serve as supplementary documentation (00:32:45).
• Legal Frameworks for AI in Policing: Lili Kazemi highlighted the importance of legislative progress, specifically citing California and Utah as frontrunners in enacting laws regarding AI transparency (00:34:03). Chris Ryan praised the California law, which mandates disclosure when AI is used to complete reports and requires that every iteration of the AI process be maintained and auditable, serving as a model for good governance (00:36:19).
• Audience Q&A on Dictation Timeframes: During the Q&A, P Murray asked about potential time parameters for dictating reports at a crime scene. Chris Ryan clarified that while there are no current legal requirements mandating a specific timeframe for dictation, it is best practice for officers to dictate their reports immediately upon returning to their vehicles while details are fresh, rather than waiting up to 48 hours (00:38:54).
• The Federal Gap in AI Regulation: Lili Kazemi and Chris Ryan discussed the current lack of a consistent federal framework in the United States regarding AI admissibility in court, noting that existing laws like the Federal Rules of Evidence are being tested by the rapid evolution of AI (00:41:07). They discussed the ongoing copyright litigation concerning large language models and the difficulty of applying current laws to generative AI (00:42:38).
• AI-Enabled Fraud and CALEA Parallels: Chris Ryan detailed the surge in AI-enabled fraud, noting that in 2025, there were over 22,000 complaints resulting in nearly $1 billion in losses, including a specific case where an employee transferred $25.6 million due to a deepfake-assisted social engineering attack (00:45:10). Chris Ryan suggested that existing frameworks like the Communications Assistance for Law Enforcement Act (CALEA) and the Report Act could be amended to require AI providers to cooperate with law enforcement and maintain records, similar to how ISPs and telephone providers currently operate (00:46:49).
• AI Culpability and Safety: Lili Kazemi raised concerns regarding AI companion apps and legal culpability when AI interactions lead to harm, such as in cases involving young people (00:49:42). They referenced a murder investigation at the University of South Florida where the suspect used ChatGPT to plan and execute a crime, leading the Florida Attorney General to consider investigations into whether AI companies should be held as co-conspirators (00:50:59).
• Future of AI Forensics: Chris Ryan and Lili Kazemi concluded by discussing the complexity of interpreting AI-generated data, such as logs and tokenization, which requires specialized skills (00:53:32). Chris Ryan drew an analogy to the evolution of cryptocurrency investigations, which began with individual officers self-educating and eventually led to the development of specialized analytical tools like Chainalysis; they predicted a similar trajectory for AI forensics where third-party technology providers and specialized training will be essential for law enforcement (00:56:02).
• Business Development and Growth Strategy: Chris Ryan explains that their growth and development roadmap is rooted in an end-user centric approach, where they engage directly with detectives to solve specific problems rather than relying on standard Requests for Proposals. Because the team consists of practitioners who have performed the job, they prioritize the needs of the public and law enforcement officers over profit-seeking (01:00:14). They maintain that their applications apply across the spectrum, from local to federal law enforcement, and even across different industries, such as school districts, where workers face similar burdens of documentation (01:01:15).
• Company Identity and Industry Scope: Chris Ryan describes their company not merely as a technology firm, but as a workforce consulting firm that provides a technology product. They note that there are approximately 18,000 law enforcement agencies in the country, many of which are small, rural organizations, and the company focuses on providing these entities with necessary resources (01:03:29).
• Overcoming Resistance to Artificial Intelligence: Chris Ryan identifies the education of administrators as the primary challenge in getting organizations to adopt their tools. They emphasize that their philosophy is built on creating safeguards and governance that align with regulatory and high-risk standards for the criminal justice system, which helps alleviate concerns regarding the use of Artificial Intelligence (01:04:24).
• Bootstrapping and Grant Funding: Chris Ryan confirms that the company is bootstrapped and grew organically, allowing them to avoid the constraints of venture capital dynamics often seen in government technology. Rahul Singh notes that they are exploring alternative funding, such as grants from foundations, which could be a viable path for the company to further research regarding Artificial Intelligence safety and governance (01:05:41).
• Defining Artificial Intelligence Native Companies: Chris Ryan and Rahul Singh discuss the term "Artificial Intelligence native," expressing skepticism that it has become a buzzword (01:08:05). Chris Ryan emphasizes that while they do leverage Artificial Intelligence, they are primarily focused on workflow optimization and helping people do their work better, rather than simply being a company that exists to deploy Artificial Intelligence (01:06:54). They argue that the market is currently experiencing over-investment and "Artificial Intelligence slop," and that companies with solid, established architecture and products are the ones likely to remain sustainable (01:08:54).
• Value Proposition and Efficiency: Rahul Singh notes that Artificial Intelligence companies succeed by adding value through eliminating waste, specifically by saving time within public service workflows (01:08:54). Chris Ryan highlights a case study involving a 24-person sheriff's office in Utah, where their platform saved 5,000 hours, effectively adding the value of more than two extra officers to the organization. They emphasize that the technology is a cost-effective alternative to hiring additional full-time employees (01:10:04).
• Technical Infrastructure and Token Consumption: Rahul Singh and Chris Ryan discuss the importance of specialized knowledge alongside technology, noting that companies that solely burn tokens to deploy generic Artificial Intelligence models are less effective (01:11:09). Chris Ryan explains that their business model does not charge for consumption, allowing them to provide a high-quality product without the burden of excessive token costs (01:12:11). They add that they do not always need frontier models for tasks like word processing and prioritize selecting the right model for the specific process (01:13:12).
• Iterative Product Development and Compliance: Chris Ryan details their iterative, module-based development process which relies heavily on user feedback. Due to the nature of their work in the criminal justice sector, they strictly adhere to Criminal Justice Information Services requirements for security and privacy, meaning they do not access the actual output data of their clients (01:14:03).
• Model Development and Future Legal Concerns: Chris Ryan states their goal is to create a proprietary law enforcement model rather than relying solely on open-source or external models like Deep Seek (01:15:09). They express concern that as technology advances, containerized models may become hosted on local servers, which could create complex legal ramifications regarding subpoenas and access if those models fall outside centralized control (01:16:15).
• Shifting Landscape of Hardware and Hosting: Rahul Singh points to the increasing availability and decreasing cost of consumer-grade hardware, such as machines using advanced chips like the Nvidia Blackwell, which enables more localized Artificial Intelligence applications (01:18:22). Regarding their platform, Chris Ryan notes that while they can support on-premise hosting, government entities have become more open to cloud-based solutions using providers like Azure and Amazon Web Services, though this trend may shift as hardware infrastructure becomes easier to manage (01:19:25).
• Civil Liberties and Legal Implications: Lili Kazemi highlights that as Artificial Intelligence tools become more integrated, civil liberties and Fourth Amendment concerns—such as protection against unreasonable searches and seizures—must be considered (01:20:22). Lili Kazemi notes that inquiries made to an Artificial Intelligence system are generally not covered by attorney-client privilege, meaning users must be aware that their digital footprint remains at risk, which adds to the legal complexities of using these tools (01:21:24).
• Future Collaboration and Conclusion: Lili Kazemi and Chris Ryan share plans to collaborate on future thought leadership regarding the topics discussed (01:22:34). Rahul Singh encourages other participants to share their stories of how they are utilizing Artificial Intelligence in business and government to help educate others. The meeting concludes with the participants expressing their appreciation for the detailed discussion and the value of sharing these case studies for future entrepreneurs (01:23:36).