Skip to content
VELKRON

PRODUCT

Verification as
a pipeline.

Output goes in. A verdict, its citations and a sealed receipt come out. Everything else is detail.

01HOW IT WORKS
S1Submit

A promotion, a decision, a document. Anything your AI produced that a regulator could one day ask about. Sent via dashboard or API.

S2Check

A model reads the artefact against the encoded rulebook and drafts findings. Code then enforces the grounding: every rule id and every quoted excerpt is verified against the rulebook and your artefact, not just prompted for. Anything unverifiable is dropped before it reaches you.

S3Verdict

PASS with the binding rules cited. FAIL with the exact breach named. Or REFUSED, when no rule matches with confidence, escalated to a human.

S4Receipt

The whole run is sealed into a hash-chained receipt and filed as an evidence pack. Months later, in an audit, you can prove what ran and what it found.

02THE THREE VERDICTS

Pass. Fail. Refuse.

Most compliance tools only know how to say yes. Velkron has three answers, and the third one is the point: when the rulebook does not support a verdict, the system says so and hands the case to a human. A confident wrong answer is worse than none.

Watch the replay: one clean promotion, one breach, one refusal.

EVIDENCE PACK · FIN-PROM-4471RUNNING
artefactpromotion_draft_v3.pdf
rule setFCA COBS 4 · FSMA s21
check 1/3risk warning present · COBS 4.2.1R✓
check 2/3balanced view of benefits · COBS 4.2.1R✓
LIVE REPLAY · 3 CASES
03THE STACK

Three products, one layer.

Promotions checkerPILOT OPEN
Checks financial promotions against a regulator's rulebook before publication. First rulebook live: UK FCA financial promotions (COBS 4, including the COBS 4.12A crypto-marketing regime, plus FSMA s21 and the FG24/1 standalone social-media guidance). The engine is rulebook-agnostic: ADGM FSRA and DIFC DFSA are next. Sold per seat plus per check, priced against the cost of one compliance officer hour. Try it at console.velkron.xyz →
Decision receiptsIN DEVELOPMENT
A tamper-evident record of any AI decision your firm makes: model, input, output, sealed in a hash chain. Works with any AI system, not just ours. If anyone edits the history, the chain breaks in plain sight. This is the general audit trail the promotions checker already writes for itself, offered as infrastructure.
Proof of inferenceRESEARCH
Zero-knowledge proof that a specific model produced a specific decision, without exposing the model weights or the customer data. Receipts prove a record was not altered; proof of inference proves the computation itself. This is the long-term moat, in active development.

See it against your own promotions.

Request a pilot›