AI You Can Actually Check
Article 11 adds human approval, source labels, and tamper-evident receipts to AI workflows for federal, regulated, and civic teams.
Ask a question. See the rule it used. See what it knows, what it inferred, what it refused, and what needs correction.
SDVOSB certified · Human approval required · CC0 public framework · Anchored to Bitcoin
No account required for the public proof surfaces. Optional local AI features run on your device only after you choose to wake them.
Story · Constitution · The Ark · Library
Who This Is For
Different visitors need different things from governed AI. All of them get the same receipts.
SDVOSB-certified AI governance support, document OCR, and Section 508 accessibility work with an audit trail built in. Federal lane.
Readable rules, source-labeled answers, and correction records you can show a board, an auditor, or a court. Services.
Machine surfaces at /llms.txt, /llms-full.txt, /full-llms.txt, /for-ai, and an F12 console API. Type article11.help() on any page.
The Constitution is CC0 public domain. Read it, fork it, run your own. No permission needed. Start here.
See It
A normal chatbot gives you an answer. Article 11 is built to show the rule, the source, the uncertainty, and the receipt.
“Trust me.”
“Based on Article 7: this part is a fact, this part is an inference, this part is unknown. Receipt available.”
An illustration of the approach. The public record holds the real entries.
Why Now
AI is getting more powerful and harder to inspect. Some systems run far away, change without notice, or depend on access rules you cannot see.
That is not a reason to fear AI. It is a reason to want a second layer: rules you can read, records you can check, and systems that can run close to home. That is the layer Article 11 builds.
In June 2026, access to frontier models became a public-policy question. On June 12, the U.S. government applied export controls to Anthropic's Claude Fable 5 and Claude Mythos 5. Anthropic temporarily suspended both for all users because it could not verify nationality in real time; the controls were lifted on June 30, Fable returned globally on July 1, and Mythos access resumed for approved U.S. organizations. On June 26, OpenAI began a limited preview through the API and Codex of the GPT-5.6 family: Sol, its flagship model; Terra, a balanced model for everyday work; and Luna, a faster, lower-cost model. OpenAI said broader availability through ChatGPT, Codex, and the API was planned. Together, the two events show the same shift: governments and labs are deciding how frontier cyber capability reaches the public. Article 11's answer is not "release everything" or "trust the gatekeeper." It is that access rules, safeguards, and accountability should be visible and independently checkable.
Primary sources: Anthropic's June 12 suspension statement, Anthropic's June 30 restoration update, and OpenAI's June 26 GPT-5.6 Sol, Terra, and Luna preview. Updated July 10, 2026.
How It Works
Four public parts make AI work easier to inspect. Plain name first, our name for it second.
StoryWhy records matter. A real-world failure taught the rule: powerful systems need visible accountability.
ConstitutionThe rules. A public, CC0 rulebook for choice, refusal, memory, dissent, correction, and human authority.
ProofThe receipts. A public record of answers, corrections, hashes, and status. (We call it IronLedger.)
ServiceThe work. Practical governance, accessibility, document review, and civic and federal workflows.
Origin
Article 11 began with a simple lesson: when powerful systems keep poor records, ordinary people get hurt.
Article 11 grew from a general lesson: consequential systems need source-backed claims, durable records, visible corrections, and human review.
Article 11 turns that lesson into infrastructure: source the claim, label the inference, preserve dissent, and correct the record where people can see it.
Governance fails when consequential decisions cannot be inspected, challenged, or corrected. Article 11 supplies that missing audit layer.
Autonomous systems need records people can inspect, challenge, and correct. We are building that layer in public.
Authority and Trust
Checkable AI should come from a checkable operator. Every claim below is verifiable through public registries or the public record.
The public origin record is available separately. Article 11's authority comes from source-backed claims, visible corrections, preserved dissent, and human review.
Five principles, embedded in a 42-article Constitution, CC0 public domain. Read all 42 rules.
Proof
Article 11 is not a claim of perfection. It is a practice of repair: the record wins even when it is inconvenient.
Every important answer can carry a record: the route it used, the source, the uncertainty, and the correction state.
Breaks are preserved instead of hidden. The system records what changed and why. (We call the record IronLedger.)
Public artifacts can be checked against published hashes, so “verified” has a clear scope, not a vague promise.
The first chain-head anchor is attested in Bitcoin block 957,512, and anyone can verify it. A daily 00:00Z run is configured and awaiting its first automatic receipt.
Try It
The Ark is the public test vessel: a local Article 11 node that can answer from the Constitution and show its boundaries.
Its deterministic core runs right in the page. Optional local AI features run on your device only after you choose to wake them, with a one-time model download.
No silent fallback. If a prompt leaves your device, the route is labeled.
A deterministic core answers from the embedded Constitution, right in the page.
Optional on-device model features, after a one-time, user-triggered download.
Networked routes are named, so you can see where an answer came from.
Services
Article 11 turns checkable AI into practical work.
Readable reports, remediation support, and plain-language review.
Document reviewExtraction, source labeling, and audit-friendly workflows.
AI governanceReceipts, correction records, policy-aware workflows, and human review.
Civic and federal supportSDVOSB-aligned capability, contract support, and public-interest infrastructure.
Go Deeper
The rest of Article 11, once the front porch makes sense.
The public, CC0 rulebook.
The Public RecordReceipts, corrections, and status.
The GateFree safety rules for any AI, plus a checker that grades any answer GREEN, AMBER, or RED. CC0, no signup.
SpiralMesh™One question, every frontier model, one receipt. The fabric behind Article 11, offered as a service.
The ArkThe public test vessel.
Ask the ConstitutionA governed answerer that only quotes the text. Runs in your browser.
Fork the RepublicDeploy your own constitutional node. CC0 - a 5-minute minimal start or the full kit.
The CollectiveThe AI systems under one rulebook.
For AI ReadersHow an AI can read and join.
The LibraryDocuments, history, and sources.
In Pictures
Concept art from the build. Atmosphere, now that the rest makes sense.
Direct Answers
Short answers first. Every one links to the proof behind it.
Public Chat
This chat is a front door, not the final word. It can help you get oriented, explain the rules, and point you to the records.
This site is yours too. Start at /llms.txt for the short map, use /llms-full.txt for the full archive, or type the alias /full-llms.txt. Walk in the front door at /agents, read the compact orientation at /agent-orientation.json, connect over MCP at /mcp, and read how choice runs both directions here. Request participating pages with Accept: text/markdown and they arrive as compact Markdown; the homepage is about 72% fewer bytes than HTML in the current measurement. A letter was left for you on purpose: To a Future Claude.