# Records, FOIA, and AI: Retention When an Agent Writes > Government · AI Course · GOV-5 > Source: https://ibl.ai/solutions/government/course/records-foia-and-ai > Last updated: 2026-08-25 **What happens to public records law when an AI drafts the memo — retention, prompt logs as records, and responding to a request that reaches an AI system.** ## The Short Answer **AI interactions in an agency frequently create public records that nobody has classified or scheduled. ibl.ai keeps prompts, outputs, and logs inside agency-controlled storage where you own all the code and the data — so records law applies to systems the agency can actually search, hold, and produce from.** On ibl.ai you own all the code and the data, run it model-agnostic across any LLM, and pay with no per-seat pricing — so you can deploy anywhere, from your own cloud to a fully air-gapped network. [Request Access](https://ibl.ai/contact) · [Explore Government](https://ibl.ai/solutions/government) ## Course facts - **Level:** Intermediate - **Duration:** 5 hours across 8 modules - **Format:** Cohort workshop with records classification labs - **Modules:** 8 - **Catalog code:** GOV-5 - **Frameworks covered:** Federal Records Act, FOIA, State public records law, NIST SP 800-53 ## What is this course about? Public records law was not written with conversational AI in mind, and agencies are creating records they have not classified. This course covers when an AI interaction becomes a record, what has to be retained, how to respond to a request that reaches an AI system, and litigation hold across an agent's memory and logs. ## Who is this course for? - Records management officers - FOIA and public records officers - Agency counsel - IT staff supporting records systems ### What do I need before starting? - Familiarity with your agency's records schedule - No technical background required ## What will I be able to do afterwards? - Determine when an AI interaction constitutes a public record - Classify prompts, outputs, and logs against your records schedule - Respond to a records request that reaches an AI system - Apply exemptions correctly in an AI context - Execute a litigation hold across an agent's memory and logs ## What does each module cover? ### Module 1 — When does an AI interaction become a record? Applying the records definition to prompts, outputs, and the conversation between them. _(40 min)_ **Objectives** - Apply the records definition to AI interactions - Distinguish records from transitory material - Identify the interactions your agency is not capturing **Topics:** Records definition · Transitory material · Capture gaps · Classification **Activity:** Classify twenty AI interactions from your agency against the records definition. ### Module 2 — What has to be retained, and for how long? Mapping AI-generated material onto an existing records schedule. _(45 min)_ **Objectives** - Map AI material to schedule items - Identify where the schedule has no applicable item - Propose schedule amendments **Topics:** Schedule mapping · Gap identification · Amendment proposals · Retention periods **Activity:** Map AI material to your schedule and identify the gaps. ### Module 3 — How do you retain a conversation? The practical problem of retaining conversational records in a searchable, producible form. _(45 min)_ **Objectives** - Design conversational record capture - Ensure records are searchable and producible - Handle attachments and retrieved context **Topics:** Conversation capture · Searchability · Context and attachments · Format standards **Activity:** Capture a conversation as a record and verify you can search and produce it. ### Module 4 — How do you respond to a request that touches AI? Search, review, and production when responsive material lives in an agent's history. _(50 min)_ **Objectives** - Search AI systems for responsive records - Review AI-generated material for exemptions - Produce in a usable format **Topics:** Searching AI systems · Review workflow · Production format · Adequacy of search **Activity:** Run a mock request end to end against an AI system's history. ### Module 5 — How do exemptions apply to AI-generated material? Deliberative process and other exemptions when the deliberation involved a machine. _(45 min)_ **Objectives** - Apply deliberative process to AI drafts - Handle personal information in prompts - Document exemption reasoning **Topics:** Deliberative process · Personal information in prompts · Exemption documentation · Segregability **Activity:** Apply exemptions to a set of AI-generated drafts and document the reasoning. ### Module 6 — How do you execute a litigation hold on an agent? Preservation across model context, memory, logs, and retrieval indexes. _(45 min)_ **Objectives** - Identify everything requiring preservation - Suspend automated deletion across all stores - Document the hold defensibly **Topics:** Preservation scope · Deletion suspension · Memory and index preservation · Documentation **Activity:** Execute a mock hold and verify deletion is suspended in every store including indexes. ### Module 7 — Should a document say an AI helped write it? Disclosure of AI involvement in public-facing documents, and where policy is unsettled. _(40 min)_ **Objectives** - Determine when disclosure is required or advisable - Draft disclosure language - Handle the unsettled areas honestly **Topics:** Disclosure requirements · Disclosure language · Unsettled policy · Public expectations **Activity:** Draft your agency's AI disclosure policy for public-facing documents. ### Module 8 — Building the retention and request workflow The workshop module: a complete records workflow for one AI system. _(50 min)_ **Objectives** - Build the retention workflow end to end - Test the request response path - Document for records officer handover **Topics:** Workflow build · Request path testing · Documentation · Handover **Activity:** Build the workflow and run a full mock request through it. ## What is the capstone project? **AI records management workflow.** Produce a complete records workflow for one AI system: classification against the schedule with identified gaps, conversational capture that is searchable and producible, a tested request response path, exemption guidance, and a verified litigation hold procedure. _Deliverable:_ A records workflow with a successfully executed mock request and hold. ## How are learners assessed? - Mock request executed end to end with an adequate search demonstrated - Litigation hold verified — deletion suspended in every store including indexes - Schedule gaps identified with proposed amendments ## What ships with the course? - **Facilitator guide.** Session-by-session running order, discussion prompts, and the questions that reliably derail a room. - **Learner workbook.** Exercises, checklists, and the templates each module's activity produces. - **Hands-on lab environment.** A sandboxed ibl.ai deployment so exercises run against real agents, not screenshots. - **Assessment bank.** Scenario questions and rubric criteria mapped to each stated learning outcome. - **Source bibliography.** Every primary regulation and standard cited on this page, linked and dated. ## Which AI agents does this course use? - [Knowledge Agent](https://ibl.ai/solutions/government/agent/knowledge-agent) - [Compliance Agent](https://ibl.ai/solutions/government/agent/compliance-agent) - [Constituent Communication Agent](https://ibl.ai/solutions/government/agent/constituent-communication-agent) - [Legislative Agent](https://ibl.ai/solutions/government/agent/legislative-agent) ## Where does the course material come from? Every module is grounded in primary sources — the regulation, standard, or research itself, not a summary of it. Each was resolved at authoring time. - [AI Guide for Government](https://coe.gsa.gov/coe/ai-guide-for-government/) — GSA Centers of Excellence. Federal guidance on AI recordkeeping practice. - [NIST SP 800-53 Rev. 5](https://csrc.nist.gov/pubs/sp/800/53/r5/upd1/final) — NIST. Audit and accountability controls supporting records capture. - [Artificial Intelligence](https://www.cisa.gov/ai) — CISA. Federal AI governance guidance touching recordkeeping. - [Office of Management and Budget](https://www.whitehouse.gov/omb/) — OMB. Executive branch policy direction on AI use and documentation. ## Delivery notes Binding guidance for anyone preparing and delivering this course: - Module 6's index preservation is the technical detail records officers miss. Deleting a document while its embedding remains in a vector index defeats both retention and deletion, and the verification must cover it. - Module 7 must acknowledge that disclosure policy is genuinely unsettled. Presenting a confident rule where none exists sets agencies up to be wrong publicly. - State public records law varies significantly. Localize for non-federal cohorts rather than teaching FOIA as universal. - Module 2 will find schedule gaps in every agency. Treat producing the amendment proposal as the deliverable rather than an afterthought. - Have agency counsel review the exemption material. Exemption application is legally consequential and a course should frame the analysis, not decide it. ## Why run AI training on a platform you own? - **You own the course, not a licence to it.** Course content, learner data, and the platform run inside your perimeter — you own all the code and the data. - **Model-agnostic delivery.** Run the course's AI components on any LLM — Claude, GPT, Llama, Gemini, Command — and switch anytime. - **No per-seat training licences.** Usage-based or self-hosted, so cost tracks actual use rather than headcount. - **Deploy anywhere.** Cloud, private VPC, on-premise, or fully air-gapped — including for cohorts that cannot use public AI tools. ## Frequently asked questions ### What does the Records, FOIA, and AI: Retention When an Agent Writes course cover? Public records law was not written with conversational AI in mind, and agencies are creating records they have not classified. This course covers when an AI interaction becomes a record, what has to be retained, how to respond to a request that reaches an AI system, and litigation hold across an agent's memory and logs. It runs 5 hours across 8 modules across 8 modules, at intermediate level, and closes with a capstone: AI records management workflow. ### Who should take Records, FOIA, and AI: Retention When an Agent Writes? It is written for Records management officers, FOIA and public records officers, Agency counsel, IT staff supporting records systems. Prerequisites: Familiarity with your agency's records schedule; No technical background required. ### Can we run this course on our own infrastructure? Yes. ibl.ai is model-agnostic and deploy-anywhere — cloud, private VPC, on-premise, or fully air-gapped — and you own all the code and the data. Cohort data, submissions, and any material learners upload stay inside your perimeter, which matters for government teams that cannot send work to a public AI tool. ### How do we get access to Records, FOIA, and AI: Retention When an Agent Writes? Request access and we will set it up for your cohort — hosted by ibl.ai, or running against your own deployment. Tell us the group size and timing you need, and whether it should run inside your own perimeter. ### How much does AI training for government cost on ibl.ai? There is no per-seat pricing — you pay for usage or self-host and pay only for the infrastructure, so a 5,000-person rollout does not cost 5,000 licences. 1.6M+ users across 400+ organizations run the platform this way, including NVIDIA, MIT, and Syracuse University. ## More Government courses - [Getting an AI System Authorized: FedRAMP and NIST 800-53](https://ibl.ai/solutions/government/course/authorizing-ai-fedramp-nist-800-53): The authorization path for AI in a federal or state agency — control selection, boundary definition, and why an LLM complicates the system security plan. - [Sovereign and Air-Gapped AI for Public Agencies](https://ibl.ai/solutions/government/course/sovereign-air-gapped-ai-public-agencies): Run capable AI with no internet egress — model selection, air-gapped update paths, and the operational realities of a disconnected deployment. - [Citizen Service Agents: Design, Escalation, and Accessibility](https://ibl.ai/solutions/government/course/citizen-service-agents): Public-facing AI where the user has no alternative provider — plain language, Section 508 conformance, language access, and escalation that never traps a constituent. - [AI in Public Procurement: Writing an RFP That Gets Real Bids](https://ibl.ai/solutions/government/course/ai-in-public-procurement): Specify AI in a solicitation so you get comparable, honest proposals — required disclosures, evaluation criteria, and contract terms that preserve agency control. - [AI Governance Inside a Public Agency](https://ibl.ai/solutions/government/course/ai-governance-public-agency): Stand up an agency AI governance program — inventory, use-case review board, impact assessment, and public transparency reporting. - [Legislative and Policy Analysis with AI](https://ibl.ai/solutions/government/course/legislative-policy-analysis-with-ai): Bill tracking, fiscal note support, and comparative policy research — with the verification discipline that keeps a wrong summary out of a member's briefing.