# ai.txt — QConsul LLC AI usage policy # https://qconsultai.com/ai.txt # Mirror of this file: https://qconsultai.com/.well-known/ai.txt (byte-identical) # Human-readable HTML version: https://qconsultai.com/ai-policy # Machine-readable index: https://qconsultai.com/llms.txt # Last updated: 2026-08-04 ## About this document A machine-readable disclosure of how QConsul LLC uses AI in its own operations — which models, who reviews what, what is never AI-generated, and disclosure practices. Inspired by humans.txt and security.txt, adapted for the AI era. Publisher: QConsul LLC — a certified Benefit Company (BCFG, 92/100, 2026). Location: Portland, Oregon, USA. Contact: https://qconsultai.com/contact ## Models used QConsul is model- and vendor-agnostic. Model selection is governed by use-case fit and QConsul's token-minimalism discipline (https://qconsultai.com/token-minimalism). Foundation model families (direct use): - Anthropic Claude - OpenAI GPT - Google Gemini - NVIDIA Nemotron Multi-model platforms (routing handled by the platform): - Perplexity (Enterprise Pro — primary orchestration substrate) - Lovable AI-enabled collaboration, data, and delivery platforms: - GitHub (code, repositories, and AI-assisted development workflows) - AirTable HyperAgent (structured data operations and agentic workflows) - Supabase (backend database, auth, and edge functions) - Notion (knowledge base, project orchestration, and AI-assisted documentation) - Lucidchart (visual architecture, process mapping, and AI-assisted diagramming) - Karpathy Wiki (foundational AI/ML concepts and technical reference) Founder's current primary toolset: Perplexity Computer for multi-model orchestration; Claude.ai, ChatGPT, Google Gems, Google NotebookLM, or NVIDIA NeMo Agent Toolkit (with nemotron-3-nano-30b-a3b) where a single model is a better fit. ## Platform security posture and governed fleets QConsul operates two separate agent fleets — Perplexity Computer (Enterprise Pro) and Airtable Hyperagent — governed by a single cross-platform agentic steward. The steward applies NIST AI RMF, ISO/IEC 42001-aligned, and EU AI Act-aligned controls across all agent workflows and platforms. EU AI Act alignment (self-attested, not a certification or conformity assessment): QConsul treats the Act's risk tiers and transparency duties as a design input. No AI system on this site evaluates, scores, or profiles visitors, and no emotion recognition or biometric categorization is used anywhere in QConsul's operations. AI-generated artwork carries visible Content Credentials (C2PA) disclosure in line with Article 50 transparency duties, and QConsul maintains AI-literacy practice for its own operators in line with Article 4. Client-side classification and conformity assessment remain with the client's legal, compliance, and security functions; QConsul is not a law firm, auditor, notified body, or certifying authority. See https://qconsultai.com/privacy#eu-ai-act. Traceable governance records: QConsul maps AI governance artifacts to applicable frameworks (NIST AI RMF, EU AI Act, ISO/IEC 42001, organizational policy, and internal control objectives) through traceable compliance records. The published crosswalk — artifact by artifact, down to specific NIST functions, EU AI Act articles, and ISO clauses — lives at https://qconsultai.com/ai-governance. That page is the canonical reference for QConsul's governance model and for the discipline QConsul calls Responsible AI Program Management: AI programs engineered for governance from Sprint 1, not governance added after deployment. Perplexity Computer posture (vendor-disclosed, not warranted by QConsul): SOC 2 Type II, GDPR-aligned operations, encrypted data in transit and at rest, short-lived file storage with configurable retention, and “no training on enterprise data” for Perplexity-managed models. Clients must verify current posture directly with Perplexity and their own legal, security, and compliance teams. Airtable / Hyperagent posture (vendor-disclosed, not warranted by QConsul): Airtable is SOC 2 Type II and ISO/IEC 27001/27701 certified, with encryption in transit and at rest, SSO/MFA/SCIM options, and GDPR-aligned operations. Hyperagent runs on this infrastructure and adds agent-level observability and browser identity controls so each agent session is auditable when accessing the web. Clients must request Airtable’s current security and AI data policy directly and engage their own legal, security, and compliance teams for regulated workloads. ## AI Context Map # QConsul LLC AI Context Map [identity] owner = "K Michael" organization = "QConsul LLC" focus = "enterprise AI governance, ROI optimization, agentic fleet management" [authoritative_sources] supabase = "Primary operational data: governed agents, workflows, ROI metrics" notion = "Operational notes, project tracking, client-specific KB" wiki = "Conceptual stack, architectures, governance patterns" github = "Source code: MCP servers, harnesses, playbooks, infra-as-code" [supabase] instance = "supabase://" schemas = ["agents", "governance", "metrics"] note = "Treat as source of truth for fleet configuration and KPIs." [notion] workspace = "QConsul HQ" databases = ["Engagements", "Playbooks", "Agent Patterns"] note = "Use for narrative context, decisions, and up-to-date workflows." [wiki] url = "https://" note = "Use as the canonical conceptual model for human-AI partnerships, guardrails, and AI program governance." [github] org = "qconsul-llc" repos = ["agent-harness", "mcp-servers", "governed-fleet", "ai-governance-playbooks"] note = "Use README and /docs as technical source of truth; ignore experimental branches unless specified." ## Review practice Each engagement defines a RACI during the Tune-Up Start or Discovery phase identifying the client-side owner Responsible, Accountable, Consulted, and Informed for deliverable reviews and signoffs. QConsul recommends: - Human review of all client deliverables. - Where useful, a secondary AI review by a model different from the one that generated or enhanced the draft, to surface a second perspective. Internal work (research notes, data processing, code exploration, synthesis) may use higher degrees of AI assistance with periodic human audit, aligned to QConsul's three oversight modes: human-in-the-loop, human-on-the-loop, and human-over-the-loop. See: https://qconsultai.com/services#governance-first ## AI-generation transparency Founder voice and original ideas — Karen Michael's perspective, ethical stance, and creative framing are human-originated. AI may assist with drafting or refinement; the core intellectual position is the Founder's. Fully human-authored (no AI generation): the "Lambda Series: Cat Fur on the Pant Legs of Tech Giants" essays and Karen Michael's author's notes. This site was designed and written in a human-AI partnership between the Founder and Lovable's AI agent. ## Content Credentials (C2PA / CAI) Every AI-generated image published on qconsultai.com carries a visible "CR" Content Credentials tag in its top-right corner. Hover, focus, or tap the tag to see the issuer, the generative tool used, the generation date, and the human edits applied afterwards. Scope: hero and section artwork is AI-generated under human art direction. The QConsul "Q" logo mark was also AI-generated under human art direction and refined by the Founder. Photographs of real people (Founder, advisory board), certification scorecards, and QR codes are NOT AI-generated and are not tagged. Registry of record: src/data/content-credentials.ts in the site source. Standard reference: https://c2pa.org/ ## Scope and accountability Two-scope rule: - Entity-level: QConsul LLC governance is independently audited by Benefit Corporations for Good (BCFG re-audits annually). - Engagement-level: deliverable-level AI use, review, and disclosure are self-attested under the SOW and the RACI agreed with the client. QConsul does not own, audit, or warrant any third-party model or platform certifications referenced above. ## Related transparency artifacts - https://qconsultai.com/humans.txt - https://qconsultai.com/security.txt - https://qconsultai.com/llms.txt - https://qconsultai.com/.well-known/llms.txt - https://qconsultai.com/robots.txt - https://qconsultai.com/sitemap.xml ## Verifying you have the current build (check this on every route) Every HTML page carries in its , and https://qconsultai.com/version.json publishes the same buildStamp for the current deployment. If the stamp on a page you fetched does not match version.json, you are holding a cached copy of an earlier deployment — re-fetch with cache bypass before reporting that page's content. version.json also publishes a "routes" map keyed by path. Each entry carries that route's authoritative canonical, title, H1, and ordered section ids, generated from the same manifest the rendered page uses. Check the entry for the exact path you fetched: if the title, H1, canonical, or section sequence differs, the copy you hold is stale. Every route on this site has its own distinct content. A page whose canonical is https://qconsultai.com/ when you requested some other path is a stale cached copy of an old home page — not that route. Re-fetch with cache bypass. Two routes where this has happened repeatedly: - https://qconsultai.com/impact — carries the Statement of Benefit Purpose, the Sustainability Statement, the Social Justice & AI Governance Statement, and the BCFG scorecard (92/100, 2026). - https://qconsultai.com/insights — carries the full catalogue of articles, essays, field notes, videos, books, and credentials. Neither body of content appears on the home page. Current home-page sections, in order: hero, who-its-for, why-qconsul, what-we-do, start-here, testimonials, faq. The insights archive is not among them, so a home-page copy listing dozens of post titles is a stale artifact rather than the current page.