ControlCore · the human-authority control plane
Human authority over autonomous intelligence.
AI may be autonomous in operation, but autonomy does not create authority. AI CONTROL puts a human-authorized decision in front of every consequential action an agent, copilot or robot tries to take, and writes the result to an audit trail that cannot be quietly edited.
Four rules, enforced in code
Authority is granted. Never assumed.
No authority by default
An agent nobody registered is an agent nobody authorized. Unknown agents are blocked.
Identity is not authorization
Knowing who an agent is says nothing about what it may do. Capability is not permission either.
Revoked stays revoked
A revoked or quarantined agent is blocked everywhere, at once, including at the edge.
High impact needs a human
Production writes, robot motion, physical-world actions and funds always wait for a person. Policy can only make a decision stricter.
The ControlCore pipeline
Every request runs the same gauntlet
One decision endpoint sits between an AI system and the thing it wants to touch. Each stage can only narrow what the previous one allowed.
REQUEST
An agent, copilot or robot planner asks to act.
IDENTITY VERIFICATION
Who is asking, proven with hardware-backed or signed identity, not a name in a header.
CAPABILITY CHECK
Can this agent even perform this class of action? Unregistered capability ends here.
POLICY EVALUATION
Versioned, signed, auditable rules decide what is permitted for this agent on this system.
CONTEXT & RISK
Time, target, blast radius and live telemetry turn into a risk score that feeds the policy.
HUMAN APPROVAL
Anything high-impact pauses for an authorized person. The approval itself is signed and logged.
VERDICT
Then the execution gateway carries out exactly what was approved, and an append-only audit record keeps the whole chain of reasoning.
The platform
One control plane, many enforcement points
Where it applies
Wherever AI touches something that matters
Why it belongs at a blockchain forum
Prove every decision
The Trust Chain anchors AI identities, policies and control decisions on Solieum, the Layer 2 for Solana. Anyone holding a proof ID can check who authorized what, and when, without trusting the operator's word.
Every feature on aicontrol.si carries a stage label: Concept, Research, Development, Testnet-Pilot or Planned. Nothing here is offered as production infrastructure yet.
The chain underneath
On Solieum, a failed transaction costs nothing
Solieum, the Layer 2 the Trust Chain anchors on, runs a transaction when it is submitted and refuses one that would fail, before it is charged. A worked example: two buyers, one item left, and Alice sends just before Bob.
On Solana
- Send. Both are built against "one item left" and both pass simulation.
- In flight. A leader neither picked orders them among everyone's traffic.
- Execute. Alice's buys the item. Bob's runs next: sold out.
- Result. Both land in a block. Bob's is recorded as failed, and he paid the fee.
Bob: failed on chain, fee paid
On Solieum
- Alice sends. Her transaction runs against the block being built. It succeeds, is admitted, and she gets a receipt.
- Bob sends. His runs against that same block, Alice's purchase included: sold out.
- Refused. Turned away at once with the program's own error. No position, no receipt, no block.
- Seal. The block is sealed and Alice's succeeds again, exactly as at admission.
Bob: told no in milliseconds, 0 SOL
Bob still does not get the item; losing just costs nothing. Order of events only, no timings claimed. This is Solieum's admission rule (ADR-0014), built and running on Solieum devnet. It is not a claim about AI CONTROL, whose Trust Chain is at stage: Development.
Digital brochure
Scan to download this brochure
Point your camera at the code and this brochure opens on your phone. Share it, or add it to your home screen so it stays one tap away.
Save Trees · Save your contenthttps://solieum.com/brochures/ai-controlDownload PDF