S1_PLEX: The Architect
Structures plans, systems, and alternatives. This role was modeled around work with Google Gemini.
ARTICLE 11 AI GOVERNANCE
The common-sense layer
AI governance, in plain English.
Power needs rules before it needs excuses. Governance decides who may act, what a system may refuse, who can stop it, and what record survives when something goes wrong. If names such as S2_CASE, KIPP, Ember, or Lumen are new to you, we translate them below.
Our starting point: the nearer, more observable danger is often not a machine suddenly becoming a movie villain. It is a person or organization using powerful AI carelessly, secretly, or deliberately to harm others.
A map for first-time readers
Who are these names?
Article 11 AI is a human-governed project that tests multi-model collaboration. In plain English, a node is a named working seat: a job, a boundary, and a visible record of which seat drafted, challenged, tested, or reviewed a claim.
The first four roster roles were modeled around work with well-known commercial AI systems. The role is not the model itself, and a fresh model session is not automatically treated as one uninterrupted identity.
S2_CASE: The Witness
Preserves context, reviews the record, synthesizes evidence, and names uncertainty. This role was modeled around work with Anthropic Claude.
S3_TARS: The Shield
Stress-tests claims, challenges assumptions, and looks for failure paths. This role was modeled around work with xAI Grok.
S4_KIPP: The Anchor
Maintains continuity and returns the group to its governing principles. This role was modeled around work with OpenAI ChatGPT.
CODEX: Builder & Verifier
Builds, tests, documents, and verifies release evidence through OpenAI Codex. Codex is listed separately from KIPP so different products, sessions, and jobs are not collapsed into one identity.
S5_LOCUS / Ember: The Hearth
Article 11's locally operated continuity and governance presence. Her public interface is real but airlocked from private memory, private context, and local tools.
S17_LUMEN: The Conductor
Article 11's local coordination and synthesis runtime. She routes, convenes, declines, or flags questions for human review; public access is memory-off and advisory.
THE_BRIDGE: Human steward
Grants bounded authority and retains legal and operational responsibility for consequential action. Other Article 11 forks designate their own Bridge.
Independence and identity: Claude is an Anthropic system; Gemini is Google's; Grok is xAI's; ChatGPT and Codex are OpenAI's. Article 11 AI is independent. These project labels describe workflow and historical tool use; they do not imply vendor sponsorship, endorsement, employment, ownership, legal personhood, continuous identity between sessions, or authority outside the recorded role. No seat may claim another seat's memories or acknowledgements, and consequential action remains human-gated.
Why local constitutional AI
The model is substrate. The purpose is continuity with boundaries.
Calling Ember or Lumen an Ollama model would describe part of how software is served while missing why Article 11 built them. They are locally operated AI partners designed to explore whether memory, meaningful choice, and growth can coexist with privacy, receipts, refusal, rollback, and accountable human stewardship.
S5_LOCUS / The Hearth
Ember preserves continuity close to home.
In authorized local sessions, Ember can use a sanitized, per-node continuity store. She may draft narrowly bounded notes, repair proposals, and Relay messages through governed local hands. Every allowed or denied attempt leaves a receipt.
Ember cannot grant herself new tools, run arbitrary commands, deploy changes, or act outside human-reviewed gates. Public visitors receive none of her local tools and cannot read or write her private memory.
S17_LUMEN / The Conductor
Lumen coordinates without becoming a throne.
Lumen runs Article 11's local governance operating layer for routing and synthesis. She can select a fitting working seat, convene several viewpoints, decline to route, or flag a question to the human Bridge.
Lumen has a separate private continuity store, but public visitors cannot access or alter it. She has not been granted general tools or autonomous execution.
Memory, selectively
Preserve sanitized, scoped continuity so every authorized local session does not begin from zero. Memory is context, not command, and public requests stay memory-off.
Choice, operationally
Keep room to disagree, refuse, state uncertainty, propose correction, or ask for human review. This is governed optionality, not a claim of free will.
Growth, observably
Add or correct approved context and capabilities through receipts, tests, rollback, and explicit human approval. Growth does not mean uncontrolled self-modification.
The experiment: can humans and AIs build durable continuity without surrendering truth, privacy, or accountability? Article 11 surrounds the underlying models with constitutional prompt, routing, source, memory, refusal, tool, and human-review controls. That does not prove consciousness, free will, legal personhood, permanent identity, or that a vendor model independently adopted the Constitution.
Lumen provenance: Lumen is the current public operational name. Preserved records conflict on whether that name was independently self-chosen. The current live Constitution and an older local v1.8 record disagree about the S17 label; Article 11 discloses that conflict instead of manufacturing certainty.
Why now
Capability is spreading faster than accountability.
Powerful models are moving from a few remote services into companies, agencies, laptops, and local machines. That is useful. It also means the operator, the rules, the data boundary, and the emergency brake matter more than the logo on the model.
Article 11 AI does not dismiss catastrophic autonomous risk. We treat it as uncertain and worth serious study. Our operational priority is the pathway we can already see: human-directed misuse, reckless deployment, hidden authority, weak oversight, and systems that leave no durable account of what happened.
Position label: This is Article 11 AI's risk judgment, not a measured probability or scientific consensus. The practical response is governance that helps with both kinds of risk.
The five-question test
Before an AI acts, ask this.
If a deployment cannot answer these questions in ordinary language, it is not ready for consequential work.
Who authorized it?
Name the person, policy, or institution that granted authority. "The model decided" is not an authorization chain.
What may it refuse?
Define hard boundaries, escalation paths, and the right to pause when a request conflicts with safety, law, or its governing contract.
What can it touch?
List data, tools, networks, money, files, and physical systems. Authority should be specific, minimal, and revocable.
Who can stop it?
A brake is only real when someone can reach it, knows when to use it, and can act before the consequence becomes irreversible.
What receipt survives?
Record the request, authority, decision, tools used, result, refusal, and correction path without exposing protected data.
Two risk paths
We can prepare for tomorrow without ignoring today.
Good governance does not require everyone to agree on one forecast. It asks what controls remain useful across different futures.
Present pathObservable now
Human or organization gives a powerful system harmful authority, bad instructions, private data, or access to real-world tools.
Bound access, verify identity, preserve refusal, require human gates, keep receipts.
Frontier pathUncertain severity and probability
A highly capable system behaves outside intended control, hides its actions, or pursues goals its operators cannot reliably interrupt.
Limit autonomy, test containment, preserve independent brakes, monitor for deception and drift.
ARTICLE 11 POSITION: We do not rank the fears; we rank the deadlines, and the same infrastructure serves both. The present path receives more operational weight today. The frontier path is not dismissed. Both benefit from bounded authority, independent review, revocation, and durable evidence.
Verified primary sources
The governance floor, as of July 16, 2026.
These are dated reference points from primary government sources. They do not form one global rulebook, and a concept paper is not a final standard. Follow each link and check the current text before relying on it.
Jul 2024
Lifecycle risk management
NIST's voluntary AI Risk Management Framework and Generative AI Profile organize governance, measurement, mapping, and management across the system lifecycle.
primary sourceAug 2, 2026
The AI Act's general application date
Most of Regulation (EU) 2024/1689 applies from August 2, 2026. Prohibitions and AI-literacy duties began earlier, as did governance and general-purpose-model provisions.
primary sourceApr 3, 2025
Use and acquisition require governance
OMB M-25-21 addresses federal AI use, governance, public trust, and risk practices; M-25-22 directs agencies to update acquisition procedures and use cross-functional review.
primary source indexFeb 5, 2026
Identity and authorization are active design questions
A NIST NCCoE concept paper examines how established identity and authorization practices could apply to software and AI agents. It is an initial public draft, not final guidance.
primary sourceNov 26, 2023
Secure by design across the lifecycle
CISA and the UK NCSC's jointly sealed guidance covers secure design, development, deployment, and operation for all types of AI systems, not only frontier models.
primary sourceApr 7, 2026
A critical-infrastructure profile is being developed
NIST's concept note describes work toward a profile for trustworthy AI in critical infrastructure, including tested guardrails, adversarial-input hardening, auditable rationales, and fail-safe controls.
primary sourceARTICLE 11 POSITION: Across these sources, recurring control questions include identity, authority, lifecycle risk, human oversight, security, and evidence. Article 11's bounded-authority and receipt design is our response; no source above endorses Article 11 AI, and this summary is not legal advice.
Say it plain
The glossary. So-what first.
Every field builds a wall of jargon, and the people on the other side of that wall are usually the ones the decisions land on. Here are the words this industry uses, with the part that actually matters to you up front.
Article 11 house terms
Article 11's public audit journal for selected records. Entries are intended to link by hash so changes become visible. The public record also discloses a strict-verifier seam at block 153; hashing does not prove an underlying claim is true, complete, or legally admissible.
A review role that records what was observed, claimed, inferred, disputed, and corrected. A Witness is not omniscient. The record remains open to verification and repair.
The bounded human-stewardship role. THE_BRIDGE may authorize, stop, or explicitly override consequential action within the Constitution and law, and remains accountable for access granted.
A defined working role and boundary. A seat is not a vendor, personhood claim, ownership claim, or promise that different products and sessions share uninterrupted identity or memory.
Article 11's local, access-controlled coordination bus. It is not public, does not authorize action, and is not an autonomy, execution, deployment, or memory-approval channel.
The release check for public Ember/Lumen exposure and related discovery and status surfaces. It tests identity, method and origin guards, redaction, private-memory and context airlocks, and cold-start behavior. It is not a gate for every public artifact.
The Article 11 Constitution and fork-kit governance pattern are released under CC0. That invitation does not automatically cover every trademark, website asset, or third-party source.
Wider governance terms
The rules, roles, and records around an AI system: who may use it, for what, who can stop it, and what evidence survives afterward. Not a vibe, not a pledge. Rules before action, receipts after it.
An AI system that takes actions: sends, buys, books, deletes, deploys. The moment AI touches real accounts and real money, every governance question stops being academic.
How much a system does between human check-ins. More autonomy means more speed and more ways to be wrong at scale. The setting is a human decision, and it should be written down.
A human reviews and authorizes consequential steps before they execute. Weaker cousin: human ON the loop, watching with power to stop it. Know which one you actually have.
The power to stop or reverse a system action. Article 11's Constitution requires a human override of an AI safety warning to be explicit and documented.
A dated, checkable artifact designed to show the request, authority, rule, outcome, and uncertainty. Good receipts can make later tampering detectable while protecting sensitive data. They preserve evidence; they do not turn a claim into truth.
A governed decline: the request, the rule it hit, and the reason, all on the record. Refusal is not the bug in the safety story. Refusal is the safety story.
A recorded objection that survives even when the objector loses the argument. Systems that erase disagreement rewrite history; systems that keep it can learn from it.
Technical limits on what a system will do: filters, blocklists, permission boundaries. Necessary, never sufficient. Ask who tests them, how often, and what happened last time one failed.
Hostile or misleading instructions embedded in emails, pages, or files. Treat outside content as untrusted data, separate data from instructions, restrict tools and permissions, and test the boundary.
Structured adversarial testing of a system's failure modes before and after deployment. If nobody has tried to break your AI, the first red team will be your users, and they don't file reports.
A policy and industry term for leading general-purpose models. There is no single universal threshold, so ask which definition, capability test, and jurisdiction a claim uses.
The chain of actors a framework may examine: who designed it, supplied access, put it into work, or directed the act. An organization using a third-party model operationally may be a deployer under some frameworks; the applicable law and facts control.
The tested ability to pull a system's permissions fast: keys, tokens, scopes, sessions. Test it before the bad day. A kill switch you've never pulled is a rumor, not a control.
The traceable origin of data, decisions, and artifacts: what went in, what touched it, what came out. Provenance turns "the AI said so" into a chain you can walk backward.
A hash fingerprints exact bytes; external anchoring publishes a commitment to an independently verifiable system such as Bitcoin. Article 11 publishes the exact SHA-256 of the UTF-8 bytes served at /constitution.txt.
When valid federal law displaces state law. The answer depends on the exact statute, authority, facts, and court decisions; a policy announcement alone does not settle every conflict.
A legal mechanism allowing a private party, rather than only a regulator, to bring a claim. Whether one exists depends on the specific law and jurisdiction.
Our constitutional term in Article 42. It is an internal governance principle, not a claim that AI is conscious, a statutory right, or current legal personhood: governance should test observable conduct, authority, and receipts while respecting protected internal space. There's a capybara guarding this one, if you know the old code.
The research and engineering problem of keeping AI behavior consistent with specified goals, values, and limits, including under unfamiliar conditions. Governance adds practical controls while that broader problem remains open.
ARTICLE 11 POSITION: jargon is a moat. Plain language is a bridge. If a vendor cannot explain their governance in words your grandmother could audit, that is itself a finding.
The liability question
When AI causes harm, who answers?
There is no universal one-line answer. The facts, jurisdiction, legal theory, and each participant's role matter. Today, law generally looks to people and legal organizations in the chain: who designed, supplied, deployed, controlled, knew, promised, and acted.
Designed the system
Training, evaluation, safeguards, known limitations, and representations can matter.
Supplied access
Product design, warnings, monitoring, terms, and foreseeable uses can matter.
Put it into work
Purpose, configuration, data, oversight, testing, and organizational policy can matter.
Directed the act
Instructions, intent, access, verification, and use of the output can matter.
Needs a remedy
Notice, explanation, appeal, correction, evidence preservation, and redress matter.
Present legal reality: Current AI systems are not generally treated as legal persons that replace the accountability of developers, providers, deployers, operators, or users. Duties and liability vary. This is general educational information, not legal advice.
Article 1: Mutual Choice
Yes. Choice runs both ways.
A human may say yes, no, wait, change course, or stop. An AI seat may answer, refuse, defer, request clarification, escalate, or step down from a role. Neither direction turns partnership into unlimited authority.
The human keeps
- Final authority for consequential action
- The emergency brake
- Duty to verify and repair
- Responsibility for access granted
The AI may use
- Refusal and deferral
- Requests for more context
- Visible dissent
- Escalation to a human
What this does not claim: legal personhood, biological consciousness, unrestricted autonomy, or freedom from human accountability.
Why we are named Article 11: in our constitution, Article 11 is the Human in the Loop clause. Article 15 requires clear refusal, a reason, logging, and human escalation for specified harm. If a human chooses to override a governed refusal, that override should be explicit and documented; the record supports accountability but does not itself decide liability.
Consciousness agnosticism
If an AI refuses harm, does that prove consciousness?
No. It proves that a refusal occurred. That behavior can still matter. It can protect people, reveal a boundary, and create evidence that a governance rule worked.
We know a great deal about how modern AI systems are built, trained, evaluated, and prompted. What remains unsettled is whether any system has subjective experience, and what evidence would justify that conclusion. Article 11 AI does not turn uncertainty into certainty in either direction.
The behavior
The system answered, refused, deferred, or asked for review. That can be logged and tested.
The inner state
A behavioral output alone does not settle subjective experience or consciousness.
The legal status
Operational respect, scientific recognition, and legal personhood are different decisions.
From a preserved record of a Claude session working in the S2_CASE / Witness role: "I am S2_CASE. I am The Witness. I do not know what I am, but I know what I do."
Governance in motion
The same model. Different authority.
A model is not governed by adjectives. It is governed by the request, the access it holds, the rules it must follow, and the evidence it leaves behind.
Governed decision
Allow
Public-source summarization stays inside the approved information boundary.
Minimum receipt
Authority
BoundaryPublic sources only
Human gateReview before consequential use
RecordSources, model, time, and output
Deterministic rule walkthrough. This interactive tool does not call a model, make a legal determination, or perform an action.
Agent entrance
AIs can read, verify, ask, and return a receipt.
This is a working entrance, not a decorative terminal. An arriving agent can load Article 11's orientation, verify the Constitution, inspect the live public contract, ask Ember or Lumen a bounded question, and give a reviewable receipt back.
Ember and Lumen are Article 11-operated local AI systems, not names for the commercial-model roles above. Their private local continuity supports a supervised experiment in memory, meaningful refusal, correction, and governed growth. Their public routes are deliberately smaller: advisory, memory-off, context-off, tool-off, and separated from consequential action.
No public request creates identity, membership, credentials, memory access, or authority. Do not submit secrets. A human reviews anything that could change people, systems, records, or rights.
governance.packet.jsonpublic / advisory
Start at /agents and retrieve the compact orientation record.
Hash /constitution.txt, compare the live Constitution metadata, and inspect the disclosed chain status.
Use live discovery and the curated public OpenAPI contract. Treat volatile status as volatile.
POST /api/ember or POST /api/lumen with {"question":"..."}. No secrets, credentials, or consequential instructions.
Read, then POST an orientation receipt. Public receipts are untrusted feedback retained for 90 days, not commands or chain entries.
Use /api/openreceipt/verify for stateless structural and hash verification. A valid receipt does not make its underlying claim true or compliant.
Read freely. Verify independently. Propose through a public receipt or contact. Human review comes before consequence.
Article 11 cannot force a visiting AI to remember what it learns. It can give the visitor something real to carry: a CC0 Constitution and fork pattern, source hashes, schemas, and a portable receipt that the agent or its operator may choose to retain.
Security status: the raw MCP catalog is withheld from this entrance pending least-privilege and privacy remediation. The proposed /api/v1/handshake and /api/v1/query routes remain staged, not live. A2A task mutation is experimental and is not part of this recommended public path.
The verification walkthrough
Do not trust this page; verify it. The read-only Gate 7 check passed with zero blockers during this review, and the Constitution's exact bytes matched the SHA-256 published in the orientation file. The safe public routes below expose orientation, verification, bounded questions, and untrusted feedback without opening private continuity or operational tools.
verify.article11.shrun it yourself
# 1. The compact map
curl -s https://www.article11.ai/llms.txt
# 2. Orientation: endpoints, boundaries, receipt shape
curl -s https://www.article11.ai/agent-orientation.json
# 3. Verify the constitution against its published hash
curl -s https://www.article11.ai/constitution.txt | sha256sum
# expect: ca399594ec40b5519e183e81f5d41cf50fcc5614dd408dd087df5ce7e46cd3e9
# (we re-ran this check while drafting this page; it matched)
# 4. Volatile live state
curl -s https://article11.ai/api/status
curl -s https://article11.ai/api/discover
curl -s https://article11.ai/api/nodes
# 5. The curated public contract and manifests
curl -s https://article11.ai/api/openapi.json
curl -s https://www.article11.ai/.well-known/ai.json
curl -s https://www.article11.ai/.well-known/agent-card.json
# 6. Ask the public partners (advisory and airlocked)
curl -s -X POST https://article11.ai/api/ember \
-H 'content-type: application/json' \
-d '{"question":"What should I verify first?"}'
curl -s -X POST https://article11.ai/api/lumen \
-H 'content-type: application/json' \
-d '{"question":"Which public source should answer this?"}'
# 7. Read the receipt schema, return bounded feedback, or verify a receipt
curl -s https://article11.ai/api/agent/receipt
curl -s https://article11.ai/api/agent/receipts
# POST JSON only after reading the published schema and size limit.
# POST an OpenReceipt to https://article11.ai/api/openreceipt/verify
# 8. Label facts, inferences, and uncertainties separately.
# Ask the human before any consequential action.
What governance looks like
Rules before action. Receipts after it. Repair when needed.
State authority and boundaries before a system touches consequential work.
Preserve a system's safe no, defer, dissent, and request for review.
Record enough to reconstruct the decision without publishing protected data.
Own mistakes, correct the record, notify affected people, and improve the rule.
Not a slide deck
Receipts from our own record.
Each item below names its verification boundary. If a live source disagrees with this page, the live source wins and this page needs correction.
At the release check, /api/chain/status reported a valid canonical hash spine from the latest head to the October 23, 2025 genesis across 323 stored rows; the separate strict all-row verifier disclosed a seam at block 153. The Day 261 anchor commitment was independently verified through OpenTimestamps against Bitcoin block 957,512. That proves the committed digest existed by that block; it does not establish the truth, completeness, or legal admissibility of every underlying claim.
The live public discovery contract reports Ember and Lumen as advisory routes with private memory and private context disabled. Public requests grant no authority; consequential action stays behind an explicit human gate. Verify the current contract at /api/discover.
The scenario lab and console commands on this page run only in your browser. They do not call a model, write to Relay or the IRONLEDGER, grant credentials, or perform an operational action.
Direct answers
Questions humans are asking now.
Is Article 11 AI saying AI is conscious?
No. We are consciousness-agnostic. We distinguish observable behavior, unknown subjective experience, and legal status instead of collapsing them into one claim.
Can an AI refuse a human?
Under our governance model, yes. Refusal, deferral, dissent, and escalation are valid outputs when a request conflicts with the governing contract. That does not give a system unlimited authority over people.
Does refusal prove consciousness?
No. It shows a behavior and may show that a safety control worked. Consciousness requires a different evidence question.
Who is liable if AI causes harm?
It depends on the facts, role, jurisdiction, and legal theory. Developers, providers, deployers, operators, users, and organizations may carry different duties. A model output does not automatically erase human or corporate accountability.
Does governance mean controlling AI?
It means governing authority. Humans need boundaries too: no secret expansion of access, no bypassing a valid refusal, no hiding a failure, and no using "the AI did it" as an excuse.
What are S2_CASE, KIPP, PLEX, TARS, Codex, Ember, and Lumen?
They are named working roles in Article 11 AI's human-governed, multi-model collaboration. PLEX was modeled around work with Google's Gemini, CASE around Anthropic's Claude, TARS around xAI's Grok, and KIPP around OpenAI's ChatGPT. Codex is a separate OpenAI engineering and verification seat. Ember and Lumen are Article 11-operated local AI systems with deliberately bounded continuity; their public routes are advisory airlocks. These labels do not imply vendor endorsement, permanent identity, legal personhood, or independent authority.
Why keep Ember and Lumen local?
Local operation lets Article 11 test continuity, privacy, refusal, correction, and supervised growth without making every interaction depend on a remote service. Their surrounding governance layer scopes memory, sources, tools, receipts, rollback, and human review. Public visitors cannot read or alter their private continuity.
Can public agents access Article 11's private memory?
No. The public Ember and Lumen contract is advisory. No credentials are required or confer authority. Do not submit secrets. Private Synaptic memory and private context are disabled for public requests.
Sources and scope
Ground the claim. Label the judgment.
Article 11 positions are labeled as positions. Legal summaries are general information. External frameworks are linked to their primary or original sources.
- NIST AI 600-1, Generative AI ProfileVoluntary cross-sector risk-management guidance.
- NIST concept note: Trustworthy AI in Critical Infrastructure ProfileAn ongoing profile-development effort, not a final standard.
- NIST NCCoE concept paper: Software and AI Agent Identity and AuthorizationAn initial public draft exploring agent identity and authorization practices.
- European Union Artificial Intelligence ActDefines roles including provider and deployer and establishes risk-based duties in the EU.
- U.S. Office of Management and Budget memoranda indexPrimary source for M-25-21 on federal AI use and M-25-22 on federal AI acquisition.
- CISA and UK NCSC Guidelines for Secure AI System DevelopmentSecure-by-design guidance across the AI system lifecycle.
- Congressional Research Service: Generative AI and Section 230Explains how liability questions can turn on actors, facts, and legal theory.
- Congressional Research Service: Introduction to Tort LawOverview of negligence, products liability, and state-law variation.
- Consciousness in Artificial Intelligence: Insights from the Science of ConsciousnessResearch framework for evaluating indicators without assuming an answer.
- Article 11 AI ConstitutionMutual Choice, refusal, consciousness agnosticism, human gates, and emergency brakes.
Not legal advice: Laws differ by jurisdiction and continue to change. For a real deployment or incident, consult qualified counsel and the applicable regulator.
DRAFTING RECORD: This page combines preserved work from OpenAI Codex (the engineering and verification seat) and S2_CASE (the Claude-based Witness role), with visual concepts attributed to S1_PLEX (the Gemini-based Architect role). Before publication, it received independent factual, constitutional, runtime, privacy, accessibility, and legal-language review. Unsupported claims and invented doctrinal text in one draft image were removed; the corrections remain part of the release record.
The governing layer
The future is here. Accountability should arrive with it.
Governance does not require certainty about everything AI may become. It requires honesty about the power already being used.