Sealby Aqta

Checked before consequence.

AI acts. Keep the receipt.

The next generation of "who cleared this?"

Logs tell you what your system recorded. Seal gives you a record an outside reviewer can verify.

Checked before the act. One file a stranger can check offline against a published key.

AI is entering regulated decisions faster than anyone can prove what it did. Any regulator asks the same question, what did the model do, on what input, under which policy, with what human oversight. Seal answers it with evidence anyone can verify.

Cross-provider

OpenAI, Anthropic, Gemini, Bedrock, Azure OpenAI. One check before every call, one receipt format. Switch providers without losing the chain.

Pre-execution

The policy decision is made before the model call. The signed receipt binds that decision and is returned with the response.

Regulator-shaped

DORA, EU AI Act, GDPR Article 30 (records of processing), EBA outsourcing. Receipts map to artefacts your auditor already knows how to read. No new vocabulary for the compliance team.

Bank-portable

Receipts verify against the published public key. Your bank's auditor can check them without calling Aqta. Independent of our uptime.

Human in the loop

EU AI Act Article 14 requires a human can intervene, override, or refuse. The session kill switch and review escalation are recorded in the audit trail today.

Sustainability disclosure

CSRD and ISSB IFRS S2 require quantitative energy and emissions disclosure. Neither names AI, so how you attribute the AI share is left to you. Each signed receipt carries an identified model, provider and timestamp. From that, we model estimated energy and carbon per decision, using token counts and standard coefficients. Estimated, not metered. A per-decision line item your assurance provider can attribute against.

Start with one model, one desk.

We attach signed receipts to every call. Tell us what you would point it at. hello@aqta.ai is read by a human, not a ticketing queue.

Now taking pilots.

A free 30-minute evidence review first, then up to five working days on one workflow: one review path, one evidence pack, judged against your reviewer’s own criteria.