The authority layer for AI agents

AI can think. It cannot give itself permission.

Capability is not authority.

Confidence is not permission.

Watch the model get stopped.

Proof, not promises.

ALEETH is the deterministic authorization and evidence layer for consequential AI-agent actions. Before an agent acts, ICA decides allow or deny, fails closed, and issues a signed receipt anyone can verify offline.

Run the live demo See the round

ConfidentialSeed 2026The Schreck Group, Inc. d/b/a ALEETH

The thesis

Capability is not authority.
Confidence is not permission.

AI agents have moved past drafting into actions that move money, change sensitive records, and start regulated workflows. When an agent can create an irreversible consequence, model confidence, retrospective logs, and governance documentation are not a control. A control decides before the action, and it cannot be talked out of the decision.

The proof

Don't take our word. Watch it, and verify the receipt yourself.

A claims agent asks to move $48,000 against a $25,000 delegated cap. ICA denies before execution and seals an Ed25519 receipt you can verify in your own browser, with no ALEETH server. It is real: the same receipts pass our public verifier on an air-gapped machine.

ICA · Authority Boundary · live, in your browserOpen full screen →

See all seven demos → Read the machine-checked proofs Open the demo full screen

The product · ICA

Institutional Control Architecture is a complete control plane, not a single gate.

Identity and delegated authority for human and machine principals
Deterministic, versioned policy decisions, the same inputs give the same result
Ed25519 signed, hash-chained receipts, verifiable offline with no ALEETH server
Machine-checked formal methods against the running decision function
A Living Certificate and framework packs that map each decision to the rule
Remediation, an agent lifecycle, and an adversarial range under attack

Proof

Machine-checked

Formal methods against the running code, not only benchmarks.

Whole plane

Not a single gate

Identity, decision, sealed record, remediation, certification, lifecycle.

Deployed

Live and certified

In production, certified by a Living Certificate, run against an adversarial range.

Why now

The rules moved in 2026. The liability did not.

Enterprise adoption of autonomous agents is running ahead of governance maturity, and independent work is now formalizing deterministic, pre-action authorization with signed records. The field is converging on the exact architecture ALEETH was built on. A validated category with no named leader gets one named by the people writing checks.

The market and the round

A validated category. A working control plane. Now, speed.

TAM

$45 to 70B

AI governance, certification, and evidence infrastructure

SAM

$6 to 12B

Serviceable across the regulated sectors

SOM

$50 to 250M

ARR, reachable with focused execution

Market figures are internal directional estimates, not a projection.

$5,000,000Sovereign Seed · SAFE

Post-money cap

$25,000,000

Pre-money

$20,000,000

To investors

~20%

A post-money SAFE, approximately 20 percent to investors before future dilution. Final terms papered with counsel.

The deck

Fifteen slides, each title a conclusion.

The full seed deck, from the authority-layer thesis through proof, traction, the market, and the close.

ALEETH · Seed Deck · 15 slidesOpen the deck →

Read the executive summary Download the deck