What is Proof-of-Silence™?
Proof-of-Silence™ is a cryptographic accountability layer for AI decisions — seatbelts for AI. It places an auditable decision boundary between an AI system proposing an action and the system actually executing it, and records every decision, including the decision not to act, as a signed, chained record.
Core principle: AI may think. AI may propose. AI may not authorize itself.
No proof → no action.
What is the current public version?
The current public implementation version is v1.1.2, frozen on 20 October 2026. Customers whose v1.1.x licence is within its twelve-month updates period receive it at no additional charge.
What verdicts does it use?
- ACT_NOW
- WAIT
- SILENCE
- REJECT
Escalation to a human is a consequence of a verdict, not a fifth verdict.
What is new in v1.1.2?
An external red/blue review of v1.1.1 found that whoever held the gate key could still record false gate results that verified, and that parameters approved by the gate could be changed in memory after the decision. No signature was forged. The verifier now rejects those records, approved parameters cannot change, and with --params it re-checks the decided amounts against the grant and the policy. Details are on the Verification page.
What came in v1.1.1?
The offline verifier can now show that a decision really went through the gate, not only that the records are intact and signed: given the grants, the gate policy and a dedicated gate key, it re-checks every ACT_NOW. Without them, VERIFIED says the gate results are the author's assertions. Details and every other fix are on the Verification page.
What came in v1.1.0?
- Action gate — before an agent acts, five checks: evidence, permission (a grant signed by a key with the grant role, never by the agent itself), policy (allowed action, allowed instruction origin, parameter caps), freshness and replay.
- Agent Firewall — deny by default for 12 actions that let an agent outgrow its job, such as buying compute, moving money, copying itself or changing its own code. Each needs a human-signed grant.
- Agent Ear — reads the chain of gate decisions and shows a timeline, the moments where the agent's behaviour changed and flags for a human to review.
Does it detect prompt injection?
No, and it does not need to. Text from a customer message, an attachment or a web page can be evidence, but it can never carry a signed grant, so it can never authorise an action on its own.
What do I receive?
Proof-of-Silence™ is supplied as a digital product. Both editions contain the same software: a Python reference implementation with the action gate, Agent Firewall and Agent Ear, an offline command-line verifier, 548 tests (including adversarial and red-team suites), 21 conformance vectors, runnable demos, example chains and documentation. The editions differ only in the licence. You can try the verifier on the real example chains before buying on the Verify before you buy page.
Is Proof-of-Silence™ software?
Yes. The Standard Edition and the Commercial Licence deliver a software package written in Python. There is no TypeScript implementation and no hosted service. The free Lite edition is an introductory framework of documents and templates, not software.
Does it stop my agent on its own?
No. Your application must call the gate before acting. Actions that never ask the gate leave no record, so pair it with an environment where the agent has no payment details, cloud credentials or write access to its own code.
Does Proof-of-Silence™ make AI decisions correct?
No. Proof-of-Silence™ does not make an AI system always decide correctly.
Cryptographic verification can demonstrate integrity within the documented trust model. It does not independently prove that every external-world input was true.
Is Proof-of-Silence™ a certificate or a legal approval?
No. Supports key EU AI Act requirements around record-keeping, traceability and human oversight. Not a certification. It is not a regulatory approval or a legal opinion.
Has its effectiveness been independently validated?
Audit status: adversarially tested by the author and QA-tested by other AI models; no independent professional security audit yet. Broader effectiveness has not been independently established.
Is the decision chain encrypted?
No. The chain is signed with Ed25519 and linked with SHA-256 — it is signed, not encrypted. Evidence is stored as hashes, so your business documents stay with you; reasons recorded with a decision, such as a rejected amount, are readable in the chain.
Does verification need the internet or our servers?
No. The verifier runs fully offline and never contacts neonai.shop. On 9 October 2026 we ran it with networking switched off: a six-record chain returned VERIFIED in under one second, including Python start-up.
Is Proof-of-Silence™ intended for medical or clinical use?
No. Proof-of-Silence™ is not designed, tested or offered for medical, clinical or healthcare decisions.
Can I get a refund?
Digital purchases are governed by the current Refund and Cancellation Policy. Statutory consumer rights remain unaffected.
For order or delivery issues contact: support@neonai.shop