<?xml version='1.0' encoding='utf-8'?>
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:video="http://www.google.com/schemas/sitemap-video/1.1"><url><loc>https://davisaigov.ai/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/about-richard-davis/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-assurance-domains/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-cyber-privacy-glossary/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-governance-labs/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-harness/</loc><lastmod>2026-11-12</lastmod><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/harness-part-1.jpg</video:thumbnail_loc><video:title>AI Harness: what it is and why it matters</video:title><video:description>An AI model can draft an answer. The surrounding system determines what it can read, remember and do. Consider a supplier invoice: summarizing it is one task; changing bank details or releasing payment is another.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/media/harness-part-1.mp4</video:content_loc><video:duration>178</video:duration><video:publication_date>2026-11-08</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/harness-part-2.jpg</video:thumbnail_loc><video:title>AI Harness: how controls and evidence work</video:title><video:description>The davisAIgov AI Harness Reference Model connects business authority to technical controls. Start with our illustrative invoice workflow. An agent reads a supplier document containing a hidden request to send customer records elsewhere.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/media/harness-part-2.mp4</video:content_loc><video:duration>314</video:duration><video:publication_date>2026-11-08</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/harness-full.jpg</video:thumbnail_loc><video:title>Complete AI Harness explainer</video:title><video:description>An AI model can draft an answer. The surrounding system determines what it can read, remember and do. Consider a supplier invoice: summarizing it is one task; changing bank details or releasing payment is another.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/media/harness-full.mp4</video:content_loc><video:duration>492</video:duration><video:publication_date>2026-11-08</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/00-generative-model.jpg</video:thumbnail_loc><video:title>AI Harness component 00: Generative Model</video:title><video:description>The engine proposes. It never holds the keys. It turns context into language, reasoning, plans, or suggested tool calls.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/00-generative-model.mp4</video:content_loc><video:duration>64</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/01-inputs-external-content.jpg</video:thumbnail_loc><video:title>AI Harness component 01: Inputs &amp; External Content</video:title><video:description>Everything that crosses into the harness: prompts, files, web pages, email, API responses, and the outputs of other tools. It matters because the agent cannot tell a fact from an instruction unless something tells it. The attack is indirect prompt injection.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/01-inputs-external-content.mp4</video:content_loc><video:duration>64</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/02-orchestrator-agent-loop.jpg</video:thumbnail_loc><video:title>AI Harness component 02: Orchestrator &amp; Agent Loop</video:title><video:description>The foreman decides who works, but the foreman is not the owner. This component selects tasks, tools, retries, and stop conditions. When a model makes one flawed recommendation, an unconstrained loop can repeat it across thousands of privileged actions before anyone notices.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/02-orchestrator-agent-loop.mp4</video:content_loc><video:duration>59</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/03-rag-subsystem.jpg</video:thumbnail_loc><video:title>AI Harness component 03: RAG Subsystem</video:title><video:description>It finds passages from your documents before the model answers. That seems like better grounding, but your document permissions have now become part of the model's behavior. A poisoned file, a stale embedding or retrieval that ignores access controls, can expose restricted information or support a wrong decision.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/03-rag-subsystem.mp4</video:content_loc><video:duration>60</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/04-memory-state.jpg</video:thumbnail_loc><video:title>AI Harness component 04: Memory &amp; State</video:title><video:description>The agent carries working context and long-term notes across steps and sessions. A sticky note left in the file can become an order for everyone who opens it later. A poisoned memory may persist silently, re-enter a decision weeks later and distort a customer case without a new malicious prompt.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/04-memory-state.mp4</video:content_loc><video:duration>61</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/05-context-assembly.jpg</video:thumbnail_loc><video:title>AI Harness component 05: Context Assembly</video:title><video:description>It combines system instructions, retrieved material, memory, user input, and task state into the model's working packet. Think of a briefing folder that quietly determines the meeting before anyone speaks. If trusted instructions and third-party text collapse into one stream, the model may treat a customer document as an order.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/05-context-assembly.mp4</video:content_loc><video:duration>61</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/06-input-output-guardrails.jpg</video:thumbnail_loc><video:title>AI Harness component 06: Input &amp; Output Guardrails</video:title><video:description>These checks inspect prompts, context, responses, and proposed actions for known hazards. They help, but a spell checker is not a signature. A filter may miss an indirect instruction or be skipped by an orchestrator that still has powerful tools.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/06-input-output-guardrails.mp4</video:content_loc><video:duration>60</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/07-policy-authorization-gate.jpg</video:thumbnail_loc><video:title>AI Harness component 07: Policy / Authorization Gate</video:title><video:description>This is the separate permission check that stands between an AI proposal and a real-world action. Sending data, spending money, changing a record. It matters because a convincing answer is not permission.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/07-policy-authorization-gate.mp4</video:content_loc><video:duration>62</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/08-credential-broker.jpg</video:thumbnail_loc><video:title>AI Harness component 08: Credential Broker</video:title><video:description>An agent needs an identity to call a database, send an email, or change a record. Treat credentials like a hotel keycard. It opens one room and expires at checkout.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/08-credential-broker.mp4</video:content_loc><video:duration>58</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/09-extensions-mcp-skills.jpg</video:thumbnail_loc><video:title>AI Harness component 09: Extensions / MCP / Skills</video:title><video:description>Every plugin is a subcontractor entering your enterprise workflow. Extensions add useful tools, external services and data sources, but they also expand what the agent can reach and execute. An unreviewed connector with broad permissions may quietly introduce vulnerable code, excessive access, or an unexpected data transfer.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/09-extensions-mcp-skills.mp4</video:content_loc><video:duration>64</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/10-tool-dispatcher.jpg</video:thumbnail_loc><video:title>AI Harness component 10: Tool Dispatcher</video:title><video:description>This is where words become side effects. The dispatcher converts an authorized request into an email, database update, payment, API call, or infrastructure change. If a malicious document persuades a model to request a destructive action, a permissive dispatcher can make the damage real.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/10-tool-dispatcher.mp4</video:content_loc><video:duration>60</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/11-sandbox-downstream-systems.jpg</video:thumbnail_loc><video:title>AI Harness component 11: Sandbox / Downstream Systems</video:title><video:description>The blast radius is set by what the room is connected to. A disposable sandbox limits where generated code and actions can go. Approved production targets need separately enforced controls.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/11-sandbox-downstream-systems.mp4</video:content_loc><video:duration>60</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/12-observability-audit.jpg</video:thumbnail_loc><video:title>AI Harness component 12: Observability &amp; Audit</video:title><video:description>The flight recorder matters more than the pilot's recollection. Your model trace, policy decision, tool invocation and actual business result must join into an independent evidence chain. If agents drift or continue work after a stop, logs written only by the agents may say everything is normal.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/12-observability-audit.mp4</video:content_loc><video:duration>63</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/13-identity-delegation.jpg</video:thumbnail_loc><video:title>AI Harness component 13: Identity &amp; Delegation</video:title><video:description>In an inter-agentic system, agents hand work to other agents, sometimes across vendors. This component keeps an accountable owner for each agent and the chain of authority between them. It matters because authority that does not narrow with each handoff is not delegation.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/13-identity-delegation.mp4</video:content_loc><video:duration>64</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/14-evaluation-conformance.jpg</video:thumbnail_loc><video:title>AI Harness component 14: Evaluation &amp; Conformance</video:title><video:description>The model name can stay the same while the behavior changes. This component links test results and continuous monitoring to the exact model, policy, prompt, tools, and retrieval configuration in use. A silent update may degrade accuracy, fairness, or refusal behavior, while the dashboard retains yesterday's passing badge.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/14-evaluation-conformance.mp4</video:content_loc><video:duration>60</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/15-human-oversight-console.jpg</video:thumbnail_loc><video:title>AI Harness component 15: Human Oversight Console</video:title><video:description>An approval badge is not oversight. A working brake is. The authorized reviewer needs the underlying evidence and a separate path to approve, reject, or interrupt a material action.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/15-human-oversight-console.mp4</video:content_loc><video:duration>59</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/16-ai-bom-attestation.jpg</video:thumbnail_loc><video:title>AI Harness component 16: AI BOM &amp; Attestation</video:title><video:description>You cannot secure an inventory you have not taken. Before an agent acts, you need to know the precise model, prompts, policies, tools, libraries, retrieval manifests, and versions assembled into its workflow. An unapproved plugin or substituted policy can slip into production while the model name remains unchanged.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/16-ai-bom-attestation.mp4</video:content_loc><video:duration>61</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/17-authority-consumption.jpg</video:thumbnail_loc><video:title>AI Harness component 17: Authority &amp; Consumption</video:title><video:description>An agent may consume calls, tokens, time and money. Every retry and child worker draws on the same business risk budget. A budget that resets at every handoff is not a budget.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/17-authority-consumption.mp4</video:content_loc><video:duration>61</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/components/18-transparency-recourse.jpg</video:thumbnail_loc><video:title>AI Harness component 18: Transparency &amp; Recourse</video:title><video:description>The person affected by an AI-assisted decision is part of the system too. A customer, patient, or employee may receive an automated outcome without a meaningful explanation or practical way to challenge an error. Raw internal traces are not a substitute for clear reasons and may disclose someone else's private information.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/components/18-transparency-recourse.mp4</video:content_loc><video:duration>61</video:duration><video:publication_date>2026-11-12</video:publication_date></video:video></url><url><loc>https://davisaigov.ai/ai-harness/ecosystem/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-harness/telemetry-evidence/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-harness/threat-map/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-regulatory-library/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-regulatory-library/energy-industrial/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-regulatory-library/financial-services-insurance/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-regulatory-library/legal-professional-services/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-regulatory-library/life-sciences-pharma/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-regulatory-library/media-technology/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/ai-regulatory-library/public-sector-critical-infrastructure/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/board-and-executive/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/capabilities/</loc><lastmod>2026-11-12</lastmod><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/capabilities.jpg</video:thumbnail_loc><video:title>Capabilities</video:title><video:description>Explore six connected disciplines: AI governance, monitoring and assurance, cyber resilience, policy defensibility, insurance and litigation support, and vendor assurance. Each scope targets a practical finding, control or evidence package.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/media/capabilities.mp4</video:content_loc><video:duration>19</video:duration><video:publication_date>2026-11-08</video:publication_date></video:video></url><url><loc>https://davisaigov.ai/clinical-risk-assurance/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/contact/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/cyber-resilience/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/cyber-resilience/harness-telemetry/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/cyber-resilience/threat-map/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/cyber-resilience/threat-map/reference.html</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/experience/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/expert-network/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/global-advisory/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/how-we-help/</loc><lastmod>2026-11-12</lastmod><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/how-we-help.jpg</video:thumbnail_loc><video:title>How we help</video:title><video:description>Choose your challenge: disputed AI actions, ransomware, regulatory scrutiny, fragmented data, rapid adoption, or deal and supplier risk. We establish facts, test controls and define next steps with clear ownership.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/media/how-we-help.mp4</video:content_loc><video:duration>19</video:duration><video:publication_date>2026-11-08</video:publication_date></video:video></url><url><loc>https://davisaigov.ai/insights/</loc><lastmod>2026-11-12</lastmod><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/insights.jpg</video:thumbnail_loc><video:title>Insights and resources</video:title><video:description>Explore our open research: the AI Harness, Trustworthy AI, Portable Trust, regulatory references, glossary, briefs and interactive labs. Follow the evidence, examine the controls and bring us your next question.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/media/insights.mp4</video:content_loc><video:duration>19</video:duration><video:publication_date>2026-11-08</video:publication_date></video:video><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/site-menu-tour.jpg</video:thumbnail_loc><video:title>Complete site menu tour</video:title><video:description>Meet davisAIgov, led by Richard E. Davis. We combine security leadership, investigation and technical evidence to clarify difficult decisions, test controls and establish accountability. Explore the experience behind our approach.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/media/site-menu-tour.mp4</video:content_loc><video:duration>113</video:duration><video:publication_date>2026-11-08</video:publication_date></video:video></url><url><loc>https://davisaigov.ai/lab.html</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/methodology/source-review/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/notices/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/portable-trust/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/privacy/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/procurement/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/request-access/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/research/openai-misalignment-disclosures/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/resilience-imperative/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/services/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/site-map/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/subscribe/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/transaction-assurance/</loc><lastmod>2026-11-12</lastmod><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/transaction-assurance.jpg</video:thumbnail_loc><video:title>Transaction assurance</video:title><video:description>Before a deal closes or a supplier is approved, examine the exposure. Our diligence supports buyers, sellers and procurement teams with prioritized risks, an evidence index and practical remediation plans.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/media/transaction-assurance.mp4</video:content_loc><video:duration>18</video:duration><video:publication_date>2026-11-08</video:publication_date></video:video></url><url><loc>https://davisaigov.ai/trustworthy-ai/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/weekly-need-to-know/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/weekly-need-to-know/2026-09-18/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/weekly-need-to-know/2026-10-02/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/weekly-need-to-know/2026-10-09/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/weekly-need-to-know/2026-10-16/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/weekly-need-to-know/2026-10-23/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/weekly-need-to-know/2026-10-30/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/weekly-need-to-know/2026-11-06/</loc><lastmod>2026-11-12</lastmod></url><url><loc>https://davisaigov.ai/who-we-are/</loc><lastmod>2026-11-12</lastmod><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/who-we-are.jpg</video:thumbnail_loc><video:title>Who we are</video:title><video:description>Meet davisAIgov, led by Richard E. Davis. We combine security leadership, investigation and technical evidence to clarify difficult decisions, test controls and establish accountability. Explore the experience behind our approach.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/media/who-we-are.mp4</video:content_loc><video:duration>19</video:duration><video:publication_date>2026-11-08</video:publication_date></video:video></url><url><loc>https://davisaigov.ai/who-we-help/</loc><lastmod>2026-11-12</lastmod><video:video><video:thumbnail_loc>https://davisaigov.ai/assets/videos/posters/who-we-help.jpg</video:thumbnail_loc><video:title>Who we help</video:title><video:description>We help boards, security leaders, deal teams, counsel, insurers, procurement and auditors across sectors. Your activities determine the relevant risks and obligations. Find the support that fits your responsibilities.</video:description><video:content_loc>https://davisaigov.ai/assets/videos/media/who-we-help.mp4</video:content_loc><video:duration>19</video:duration><video:publication_date>2026-11-08</video:publication_date></video:video></url></urlset>
