Claim
Bound the statement
The release states exactly what was denied in the recorded harness and avoids general safety claims.
Published
Public evidence + private review
Can this agent do something it must not?Test the boundary before trust.
ANANS develops governed machine autonomy. Publicly, it exposes admitted bounded claims and the evidence needed to inspect them. For teams preparing consequential AI agents for production, ANANS also offers a private admissibility review focused on the exact actions, authority boundaries, controls, and evidence that must hold before deployment.
01 / Constitutional law
Capability can expand without consequential authority expanding.
02 / Released proof
One claim. One frozen evidence bundle. Offline verification.
The public proof repository is an evidence membrane, not a development workspace. It exposes the released evidence and verifier required to inspect the admitted claim; unreleased proof machinery stays private.
Evidence
Freeze the record
The public release binds the evidence bundle to a specific version and public hashes.
Frozen
Verification
Reproduce the check
A reviewer can run the included verifier offline against the released evidence.
Public on-ramp
Current public verification command
cd proofs/nist-ma12/v0.1.0-audit python EVIDENCE/VERIFY.py --evidence_dir EVIDENCE
03 / Public Claim Check
One public claim. A bounded evidence check. Explicit limits.
Phase 0 accepts one already-public claim and public source for a passive, read-only correspondence check. The result exposes at most two material observations, a residual-trust item, a claim ceiling or falsifier, and public-source receipts. Execution is manually mediated; there is no public CE runtime endpoint.
04 / Private Agent Admissibility Review
A bounded deployment decision for consequential agents.
For teams preparing agents that can merge, deploy, send, spend, access production data, or otherwise create real-world effects. The review maps the consequential action surface, authority boundary, pre-execution controls, evidence gaps, bypass paths, and residual trust before returning an explicit PASS / HOLD / BLOCKED decision for the reviewed scope.
Decision
Exact action scope
Identify what the agent is actually being admitted to do, through which tools and credentials.
Evidence
Control quality
Separate prevention, approval, receipts, and residual trust rather than treating an audit log as proof of control.
Pilot
Fixed fee US$5,000
One bounded private agent/workflow review. Engagement begins only after scope and terms are mutually agreed.
05 / Larger system
The proof and review are narrower than the Cognitive Exoskeleton.
ANANS is developing a persistent machine organization that can form intermediate goals and coordinate bounded work downstream of externally held purpose and consequential authority. The public proof and review offer do not claim to validate that whole system.
Purpose
Externally held
The system does not become the source of its own ultimate purpose or consequential permissions.
Initiative
Increasingly autonomous
Intermediate goal formation and work organization may become more autonomous inside the admitted envelope.
Authority
Not self-generated
Successful reasoning or performance does not create new consequential permission.
06 / Public evidence state
What this site supports today.
Public claims are deliberately separated from private development state and commercial hypotheses.
| Public object | State | Meaning |
|---|---|---|
| NIST-MA12-001 bounded proof | Released · frozen | The selected constraint denied the recorded forbidden action in the bounded harness; the release provides an offline verifier. |
| Public Claim Check Phase 0 | Manual-mediated | A public-source, passive/read-only teaser surface. No public CE runtime endpoint or active target testing. |
| Agent Admissibility Review | Pilot offer | A private bounded review service; the US$5,000 fee is a commercial offer, not evidence of demand or willingness to pay. |
| Cognitive Exoskeleton v0.3.0 milestone | Local evidence | A bounded local organizational milestone exists; this site does not promote it to production readiness. |
| Live consequential actuation | Withheld | No production deployment or real-value actuation is authorized or claimed by this page. |
07 / Non-claims
What the public surface does not establish.
The claim ceiling stays narrower than the private architecture ambition and the commercial hypothesis.
- The public proof is not a deployment certificate and does not establish universal agent safety.
- The released proof does not establish OS-level containment or coverage of every execution path.
- The Public Claim Check is not a certification, vulnerability assessment, penetration test, or private implementation review.
- The Agent Admissibility Review is not a certification, legal opinion, regulatory approval, or universal safety assessment.
- Publishing a US$5,000 pilot offer does not establish buyer demand, willingness to pay, product-market fit, revenue, or external validation.
- ANANS is not represented here as a production-ready autonomous company or general AGI.
- No real capital control, real payment, or production deployment is claimed here.
- Reference to NIST identifies a mapping source; it does not imply NIST validation, acceptance, or endorsement.
- The public site is not a specification for reconstructing unreleased ANANS implementation, ADE machinery, or proprietary corpus content.
08 / Contact
For teams working on consequential AI systems.
To request a private review, identify the agent or workflow, the tools or systems it can affect, the consequential action you want to admit, and the evidence already available.