![\"Artificial](\"https://cms-nasimpson.goingblu.com/wp-content/uploads/2023/07/ai_effort_benefit_matrix-1024x1024.png\")
Every few years, the frontier of “public record” expands in ways that make lawyers nervous and IT people shake their heads. Email did it in the 90s. Text messages did it in the 2000s. Slack channels and Teams chats did it more recently. And now we’ve arrived at the next step: artificial intelligence platforms.
\n\n\n\nThere is a myth floating around government offices, sometimes whispered by staff who don’t want their rough drafts discovered, sometimes encouraged by vendors, that anything typed into an AI system is somehow outside the reach of public transparency laws. As if the content is different, or ephemeral, or “not really a document” because it came from a chatbot instead of a yellow legal pad.
\n\n\n\nVirginia law disagrees.
\n\n\n\nVirginia FOIA Cares About Substance, Not Format
\n\n\n\nThe Commonwealth’s FOIA statute has one of the broadest definitions of “public record” in the country. It is indifferent to the container, indifferent to the technology, and indifferent to how polished or unpolished the writing may be.
\n\n\n\nThe statute says:
\n\n\n\n“Public record” means any writing or recording… regardless of physical form or characteristics… prepared by, or in the possession of, a public body or its officers, employees, or agents in the transaction of public business.
— Va. Code § 2.2-3701
That definition swallows nearly every imaginable artifact of modern work. If a public employee types something into a system, any system, while conducting public business, it is a public record. Paper, PDF, Word file, text message, sticky note, Google Doc comment, Slack DM, handwritten margin scribbles, or an OpenAI “prompt history” log. FOIA doesn’t care. Medium is irrelevant.
\n\n\n\nA prompt written into ChatGPT asking it to “draft tonight’s zoning presentation” is a public record. A block of AI-generated text that becomes a briefing memo is a public record. Even the back-and-forth refinement, the entire chain of edits, prompts, clarifications, and corrections, is a public record, because the statute looks at what the employee did, not the tool they used.
\n\n\n\nThe “Drafts” and “Working Papers” Myth
\n\n\n\nSome employees assume that anything “draft-like” is automatically shielded. But Virginia is unusually strict about what counts as a “working paper.” The exemption is very narrow and applies only to the personal or deliberative use of specific officials:
\n\n\n\nThat list does not include ordinary staff, department heads, technical employees, administrative aides, or analysts.
\n\n\n\nSo, if a staff member uses AI to produce:
\n\n\n\nthose drafts are not exempt simply because they are drafts. The law treats an AI-generated draft no differently than a Microsoft Word draft saved at 2:17 p.m. on a Tuesday.
\n\n\n\nUnless the draft was created specifically for the personal deliberative use of one of the narrow officials listed above, it is not protected.
\n\n\n\nRoutine AI Use = Routine FOIA Exposure
\n\n\n\nOnce AI becomes a normal part of doing government work, it also becomes part of the public record landscape.
\n\n\n\nIf an employee uses ChatGPT, Claude, Copilot, Gemini, or anything similar to:
\n\n\n\nthe input and output are presumptively FOIA-able.
\n\n\n\nAnd not just the final answer. The entire chain: every prompt, iteration, correction, clarification, alternate version, expansion, contraction, and sample paragraph. If a person typed it in the course of public business, the public may request it.
\n\n\n\nAn AI transcript is no different than a chain of tracked changes, email revisions, or handwritten notes in a meeting notebook. It is part of the deliberative record—unless a separate exemption applies (attorney-client privilege, criminal investigative materials, personnel records, IT security documents, etc.). But the “draft” label alone does not protect it.
\n\n\n\nWhat About Logs Held by the Vendor?
\n\n\n\nThis is where the cloud era complicates things, but Virginia FOIA has been ahead of this issue for years.
\n\n\n\nAnd if they are accessible, they are subject to FOIA.
\n\n\n\nOnly records that do not exist in the possession or control of the public body are not subject to FOIA. A log stored by OpenAI that the county cannot or has no contractual right to retrieve, is not a public record. But that is far rarer than most assume, especially in enterprise agreements.
\n\n\n\nA Practical Example: The Chain of Custody Problem
\n\n\n\nImagine a zoning staffer asks an AI system:
\n\n\n\n“Draft a polite response explaining why we cannot adjust the setback requirements for Parcel #118A.”
\n\n\n\nThe AI gives a paragraph. The staffer replies:
\n\n\n\n“Make it firmer. Less apologetic.”
\n\n\n\nThe AI revises. The staffer says:
\n\n\n\n“Remove the part about last year’s case, don’t mention the Smith subdivision.”
\n\n\n\nAnd the AI produces a final version, which is emailed to the citizen.
\n\n\n\nUnder FOIA, the citizen is entitled not just to the polished email, but also the earlier versions, the internal prompts, the edits, the instructions, the “make it firmer,” the removed reference to the Smith subdivision, and the entire digital trail of how the message came to be.
\n\n\n\nThat is how the law has always worked. The AI system simply automates the drafting process that used to happen inside a Word document or inside a notebook margin.
\n\n\n\nThe Bottom Line
\n\n\n\nVirginia FOIA does not bend merely because the work is done inside a modern tool.
\n\n\n\nIf a public employee uses AI while conducting public business:
\n\n\n\nare all public records, unless a specific statutory exemption applies.
\n\n\n\nThe new world of AI doesn’t replace public accountability. It expands the paper trail, and the paper trail has always belonged to the people.
\n","protected":false},"excerpt":{"rendered":"Every few years, the frontier of “public record” expands in ways that make lawyers nervous and IT people shake their heads. Email did it in the 90s. Text messages did it in the 2000s. Slack channels and Teams chats did it more recently. And now we’ve arrived at the next step: artificial intelligence platforms. There … Read more
\n","protected":false},"author":1,"featured_media":493,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11,15,72,41],"tags":[],"class_list":["post-492","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-artificial-intelligence","category-government-technology","category-morningside","category-virginia-government"],"_links":{"self":[{"href":"https://cms.nasimpson.com/wp-json/wp/v2/posts/492","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https://cms.nasimpson.com/wp-json/wp/v2/posts"}],"about":[{"href":"https://cms.nasimpson.com/wp-json/wp/v2/types/post"}],"author":[{"embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/users/1"}],"replies":[{"embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/comments?post=492"}],"version-history":[{"count":0,"href":"https://cms.nasimpson.com/wp-json/wp/v2/posts/492/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/media/493"}],"wp:attachment":[{"href":"https://cms.nasimpson.com/wp-json/wp/v2/media?parent=492"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/categories?post=492"},{"taxonomy":"post_tag","embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/tags?post=492"}],"curies":[{"name":"wp","href":"https://api.w.org/{rel}","templated":true}]}},{"id":193,"date":"2023-07-24T11:28:04","date_gmt":"2023-07-24T15:28:04","guid":{"rendered":"https://cms-nasimpson.goingblu.com/?p=193"},"modified":"2025-10-09T07:06:57","modified_gmt":"2025-10-09T11:06:57","slug":"unleashing-the-power-of-ai-in-local-government-optimizing-document-management-and-accessibility","status":"publish","type":"post","link":"https://cms.nasimpson.com/2023/07/24/unleashing-the-power-of-ai-in-local-government-optimizing-document-management-and-accessibility/","title":{"rendered":"Unleashing the Power of AI in Local Government: Optimizing Document Management and Accessibility"},"content":{"rendered":"\nThe Role of AI in Revolutionizing Document Management
\n\n\n\nIn this continuing series on the future of artificial intelligence (AI) in the public sector, let’s explore how AI can dramatically alter the way local governments handle their document systems. As e-Government (eGOV) initiatives are gaining ground globally, the need for smarter, efficient, and more secure document management techniques are paramount. AI promises to reshape the way local governments interact with, categorize, and recover data, laying a valuable foundation for future decisions.
\n\n\n\nNatural Language Search: Empowering Interactive Data Dialogues
\n\n\n\nHistorically, government data interaction has been a one-way street, create and file away. But AI is changing the game. Cutting-edge AI technologies like Natural Language Processing (NLP) facilitate a two-way dialogue with data, enabling the computer-assisted classification and quick retrieval of records. NLP (and a slew of other AI/ML powered analysis systems) allows government workers to answer queries more promptly and provide superior public service.
\n\n\n\nEntity Recognition: Crafting Valuable Historical Records for Informed Decision-Making
\n\n\n\nAI’s entity recognition is a game-changer in enhancing data utility. This functionality detects ‘entities’ — primarily common nouns like persons, places, things, or ideas. In a government context, it’s equally important to identify events and actions. By connecting these entities, events, and actions, we can create a detailed historical record for robust future decision-making.
\n\n\n\nFor example, consider a local coffee shop. The coffee shop building is the entity, identified by its address.
\n\n\n\n1694 West Oak Street
(Entity)
The property tax payment is an event, while a zoning special use permit is also an event.
\n\n\n\nSpecial Use Permit application for 1694 West Oak Street by Lucy Graves on behalf of Oak Street Coffee, LLC
Staff approval of the special use permit is an action.
\n\n\n\nZoning Administrator Gregg Lowry Approved Special Use Permit #6409 for 1694 West Oak Street on December 12, 2017
Using natural language queries, we can learn about the history of the coffee shop and the building as it pertains to local government. These interconnected records provide us a more in-depth understanding of civic matters and insights for potential service enhancements.
\n\n\n\nImplementation: Conquering Unstructured Data with AI
\n\n\n\nLocal government agencies often grapple with integrating new software, particularly when it pertains to human-created and maintained records. But AI proves adept at overcoming this hurdle. By harnessing machine-based entity detection, NLP, and vectorization, AI can automatically ingest and classify unstructured data with high a degree of accuracy. This significantly cuts down manual labor and minimizes errors, facilitating eGOV initiatives and boosting overall efficiency.
\n\n\n\nImmutable File Storage: A Shield Against Ransom Attacks
\n\n\n\nImmutable file storage can significantly enhance the security of local government data. When records are ingested into a vector database, a ‘ledger version’ can be created in parallel to the living database. This ledger would be invulnerable to alterations. This approach could prove to be a significant countermeasure against ransom attacks that often attempt to manipulate or encrypt data for nefarious purposes.
\n\n\n\nIntegration: Advancing Data Accessibility and Archival Longevity
\n\n\n\nIntegration serves as the final linchpin. By offering the ingested data to third-party software vendors, audit and assurance partners, and the general public through standard APIs, governments can greatly improve realized value and data accessibility. Moreover, this strategy enhances archival longevity by migrating data from cumbersome legacy file formats to structured, lightweight databases. When required, legacy formats (such as Word or PDF) can be regenerated from the structured data, ensuring extensive compatibility while preserving modern data management benefits.
\n\n\n\nIn conclusion, the integration of AI in document management within local government has the potential to revolutionize data management, accessibility, and utilization. AI not only champions efficiency and precision but also fosters better public service, cementing its critical role in the advancement of eGOV initiatives. I’ll be back with more insights on the transformative power of AI, and in depth guides for implementing emerging tools in the public sector in this ongoing series.
\n","protected":false},"excerpt":{"rendered":"The Role of AI in Revolutionizing Document Management In this continuing series on the future of artificial intelligence (AI) in the public sector, let’s explore how AI can dramatically alter the way local governments handle their document systems. As e-Government (eGOV) initiatives are gaining ground globally, the need for smarter, efficient, and more secure document … Read more
\n","protected":false},"author":1,"featured_media":195,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11,15,12],"tags":[],"class_list":["post-193","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-artificial-intelligence","category-government-technology","category-writing"],"_links":{"self":[{"href":"https://cms.nasimpson.com/wp-json/wp/v2/posts/193","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https://cms.nasimpson.com/wp-json/wp/v2/posts"}],"about":[{"href":"https://cms.nasimpson.com/wp-json/wp/v2/types/post"}],"author":[{"embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/users/1"}],"replies":[{"embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/comments?post=193"}],"version-history":[{"count":0,"href":"https://cms.nasimpson.com/wp-json/wp/v2/posts/193/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/media/195"}],"wp:attachment":[{"href":"https://cms.nasimpson.com/wp-json/wp/v2/media?parent=193"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/categories?post=193"},{"taxonomy":"post_tag","embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/tags?post=193"}],"curies":[{"name":"wp","href":"https://api.w.org/{rel}","templated":true}]}},{"id":138,"date":"2023-07-03T12:37:35","date_gmt":"2023-07-03T16:37:35","guid":{"rendered":"https://cms-nasimpson.goingblu.com/?p=138"},"modified":"2025-10-09T07:06:57","modified_gmt":"2025-10-09T11:06:57","slug":"early-adoption-the-advancements-opportunities-and-future-of-large-language-models","status":"publish","type":"post","link":"https://cms.nasimpson.com/2023/07/03/early-adoption-the-advancements-opportunities-and-future-of-large-language-models/","title":{"rendered":"Early Adoption – The Advancements, Opportunities, and Future of Large Language Models"},"content":{"rendered":"\nIn the realm of artificial intelligence, the development of OpenAI’s GPT-4 has been a significant breakthrough. Large language models like GPT-4 have the power to transform industries, reshape workflows, and redefine our interaction with technology. Initially released as a chat tool, this generative model has rapidly evolved into a powerful AI that developers are now integrating into various software tools, thus opening up exciting new avenues for innovation.
\n\n\n\nLarge language models like GPT-4 use a blend of data scraping and human training. The model is fed with vast amounts of text data which it utilizes to generate relevant responses, mimicking the way humans converse, write, and think. It employs advanced machine learning, which interprets the context and relationship of words in a sentence to predict the following word.
\n\n\n\nThe exact intricacies of these models remain unknown. While model developers provide general training data, the specific documents the model has been trained on remain unknown, maintaining the privacy of data sources.
\n\n\n\nToday, GPT-4 has seen limited adoption in an enterprise setting, but hobbyists and small development teams are building complex solutions around OpenAIs GPT-4 API. It is extensively used for tasks like editing written content, natural language response generation, and document analysis. More advanced users have harnessed its capabilities to automate routine tasks by building workflows with existing SaaS tools like Google Sheets and Microsoft Sharepoint. The model’s flexibility and ease of integration have facilitated its initial adoption.
\n\n\n\nCertain areas present a perfect fit for GPT-4, where its deployment encounters minimal onboarding friction and ensures a short time to value. These areas are opportunities for high-profit margins and will be tapped into by consulting firms, who utilize their deep client trust to introduce AI solutions that would otherwise face slow adoption. This “procurement in the gaps” strategy will fuel integrating large language models like GPT-4 in various business-to-corporate and business-to-government environments.
\n\n\n\nLarge Language Models (LLMs) like GPT-4 can be implemented quickly in several ways to enhance productivity and streamline processes in both corporate and government settings. Here are a few examples:
\n\n\n\n1. Automated Report Generation: Corporate and governmental entities often need to generate regular reports. An LLM can be trained to automatically create these documents from structured data, reducing the time and effort spent on this task and freeing up human resources for more complex duties.
\n\n\n\n2. Email Screening and Response: AI can be used to filter and sort incoming emails based on their importance, sender, and content, ensuring that high-priority messages are dealt with promptly. Moreover, LLMs can be trained to draft responses for common queries, further enhancing efficiency.
\n\n\n\n3. Customer Service Chatbots: AI-driven chatbots can provide instant responses to customer inquiries, significantly reducing wait times and improving customer satisfaction. These can be deployed on various platforms, including websites and social media.
\n\n\n\n4. Document Review and Summary: LLMs can be trained to review, extract key points from, and summarize lengthy documents, reports, or legal contracts. This can be particularly helpful in the government sector, where dealing with extensive paperwork is commonplace.
\n\n\n\n5. Meeting Scheduling and Management: AI can be used to schedule meetings, send reminders, and draft meeting minutes based on transcribed audio, streamlining administrative tasks.
\n\n\n\n7. HR Processes: Routine HR tasks like screening resumes, scheduling interviews, and answering common candidate queries can be automated using LLMs, speeding up the hiring process and freeing up HR personnel for more strategic roles.
\n\n\n\n8. Social Media Management: Corporations and government entities often maintain multiple social media accounts. AI can be used to manage and schedule posts, respond to comments, and analyze user engagement and sentiment.
\n\n\n\n9. Public Opinion Analysis: Government agencies can use AI to monitor and analyze public sentiment on social platforms regarding various policies and initiatives, enabling them to make more informed decisions.
\n\n\n\nThese tasks represent the “low-hanging fruit” that can be quickly addressed by AI, offering an immediate return on investment. The real value of LLMs like GPT-4, however, will be realized as they are progressively integrated into more complex workflows, transforming processes and delivering unprecedented efficiency.
\n\n\n\nDespite the advancements, there is a gap in the seamless integration of these models. Training specific models for niche applications is expensive, and there isn’t an efficient way to teach these models company-specific information. However, it’s anticipated that these hurdles will be overcome in the near term, further broadening the possibilities for AI deployment.
\n\n\n\nAI integration into the corporate and government sectors will be spearheaded by existing software giants like Microsoft, Salesforce, IBM, and Adobe — along with the Usual Suspects of consulting firms and implementation partners. These players are positioned to augment their existing commercial-off-the-shelf tools with AI, making them more competent and efficient.
\n\n\n\nWhile existing enterprise software providers will likely be the face of AI implementation, the indie development community will drive the creation of tools that integrate GPT-4 into solutions for hyper-niche business cases, tailoring the capabilities of LLMs and AI in general to meet specific needs.
\n\n\n\nThe journey of GPT-4 from a chat tool to a robust language model integrated into various applications signals the promising future of AI. As the gaps in implementation are addressed and more players leverage AI capabilities, the landscape of technology, business, and government operations are set to undergo an incredible transformation.
\n","protected":false},"excerpt":{"rendered":"In the realm of artificial intelligence, the development of OpenAI’s GPT-4 has been a significant breakthrough. Large language models like GPT-4 have the power to transform industries, reshape workflows, and redefine our interaction with technology. Initially released as a chat tool, this generative model has rapidly evolved into a powerful AI that developers are now … Read more
\n","protected":false},"author":1,"featured_media":153,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11,12],"tags":[],"class_list":["post-138","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-artificial-intelligence","category-writing"],"_links":{"self":[{"href":"https://cms.nasimpson.com/wp-json/wp/v2/posts/138","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https://cms.nasimpson.com/wp-json/wp/v2/posts"}],"about":[{"href":"https://cms.nasimpson.com/wp-json/wp/v2/types/post"}],"author":[{"embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/users/1"}],"replies":[{"embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/comments?post=138"}],"version-history":[{"count":0,"href":"https://cms.nasimpson.com/wp-json/wp/v2/posts/138/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/media/153"}],"wp:attachment":[{"href":"https://cms.nasimpson.com/wp-json/wp/v2/media?parent=138"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/categories?post=138"},{"taxonomy":"post_tag","embeddable":true,"href":"https://cms.nasimpson.com/wp-json/wp/v2/tags?post=138"}],"curies":[{"name":"wp","href":"https://api.w.org/{rel}","templated":true}]}}]}