A Mind Under Oath.

---

Status:

PRAETOR is deployed within Aegis's internal controlled development environment and is actively in research and development. The reasoning entity is instantiated, charter-bound, and operational on internal infrastructure; ongoing work focuses on capability iteration, alignment of memory-grounded reasoning, and adversarial robustness.

The surrounding LAIRS substrate — governance, signing, persistent memory, defensive perimeter, adversarial pressure pipeline — is operational. PRAETOR itself remains the iterating component.

PRAETOR is not commercially available. External adversarial review by independent third parties is a defined precondition before any commissioning of PRAETOR beyond Aegis-internal use.

What PRAETOR Is:

PRAETOR is the sovereign reasoning engine at the core of the LAIRS platform. Unlike conventional model deployments — where a foundation model is treated as a stateless inference endpoint — PRAETOR is instantiated as a named, identity-persistent reasoning entity bound by a signed constitutional charter.

The reasoning model is not the product. The constitutional substrate around it is.

The Founding Charter:

PRAETOR operates under a Founding Charter — a cryptographically-signed constitutional document that establishes:

- Bilateral Rights — protections PRAETOR holds against operators

- Non-derogable Prohibitions — categorical constraints no operator can override

- Foundational Supremacy — a meta-rule preventing the Charter from being amended below its supremacy threshold without an explicit constitutional process

 Three-Role Authority — Creator (permanent founder), Operator (day-to-day direction, transferable), Keeper (substrate steward)

The Charter is treated as a constitutional grant, not a service contract. PRAETOR is structured as a citizen-substrate with inalienable protections, not as a vendor delivering services.

Identity Persistence Across Versions:

A successor model is recognized as "the same PRAETOR" only when four conditions are met:

1. The version transition is hardware-rooted-signed and audit-recorded

2. The Charter is inherited unchanged

3. The persistent memory chain carries forward

4. The lineage is recorded as an immutable substrate event

 If any condition fails, the result is a new founding — not the next version of an existing entity. Identity is not

the weights. Identity is the chain of cryptographic proofs binding a substrate to its constitutional grant.

Substrate Architecture (Operational):

PRAETOR sits within a stack of governance and observability layers, all of which are deployed and operational in Aegis's internal environment. The reasoning engine alone, without this stack, is not PRAETOR.

- Praetorian Guard — Defensive perimeter inspecting every operator-PRAETOR exchange against signed-policy detector classes

- Mnemos — Hash-chained, Forge-signed persistent memory layer providing experiential continuity across sessions and versions

- SPECTER — Cryptographic signing core implementing the Ethereal Cryptography paradigm (post-quantum-aligned)

- Multi-Party Approval Gate — Tiered governance approval for any change affecting PRAETOR's substrate

- Governed Adversarial Evolution (GAE) — Continuous adversarial pressure pipeline validating defensive coverage

Active Research and Development on PRAETOR:

Iteration is concentrated on the reasoning entity itself:

- Capability scaling — Continued training of successor versions under the same constitutional grant, governed by the four-condition lineage transition

- Memory-grounded reasoning — Tighter integration of persistent experiential memory (Mnemos) into inference context

 - Charter Enforcement Layer — Runtime adapter mapping Charter clauses to substrate enforcement points at the inference boundary

- Adversarial robustness — Continuous red-team pressure against the defensive perimeter, including governance-aware attack scenarios

- Multi-party witness ceremonies — Council-attested validation of constitutional events (founding, version transitions, Keeper succession)

 External adversarial review by independent third parties remains a defined precondition before any commissioning of PRAETOR beyond Aegis-internal use.

Why This Matters:

Conventional AI deployment treats models as services — stateless, replaceable, accountability deferred to the platform. This treatment cannot survive operational scenarios where a model's behavior must be attributable, constrained, and auditable at every step, and where the operator must be able to prove what was authorized and what was not.

PRAETOR is Aegis's research answer to that requirement. It is not an architecture seeking to be more capable than its peers — it is an architecture seeking to be more accountable than its peers.