The OWASP AI Exchange now cites the Asqav compliance-receipts profile. The Exchange's page on input threats, in the section on monitoring and secure logging, describes a signed-receipt approach to tamper-evident logging and links draft-marques-asqav-compliance-receipts as one open profile that sets out receipt-format obligations under the EU AI Act, DORA, and US sector regulations.
What the citation says
The secure-logging guidance recommends tamper-evident logging to verify that records stay unmodified after an incident. It describes a signed-receipt approach that binds a hash of the input, the output, and the processing context to a per-event signature, paired with an independent time anchor so the signer cannot backdate the "when". It then names the compliance-receipts profile as an open definition of those receipt-format obligations. The citation sits on the page at owaspai.org/docs/2_threats_through_use.
The Exchange also sets out why tamper-evident logging is harder for AI than for conventional applications. Output is non-deterministic, so a signature over the response alone proves little. Retention runs to multiple years under the EU AI Act and DORA. And for agents it matters who acted under whose authority, and which model and tools were available when the action was signed. Those are the properties the receipt format is built around.
Who runs the AI Exchange
The OWASP AI Exchange is an OWASP Flagship project. Its founder and lead editor is Rob van der Veer, Chief AI Officer at Software Improvement Group, who also works on ISO/IEC 27090 and the EU AI Act standard in CEN/CENELEC. Through an official liaison partnership, the Exchange feeds its content into ISO/IEC 27090, the international standard for AI security. Rob van der Veer personally thanked the founder for the contribution that led to the citation.
What it means
The citation is independent recognition of the signed-receipt approach to AI-agent accountability. The Exchange describes the same shape Asqav ships: a signature that binds what went in and what came out to the processing context, anchored so the signer cannot rewrite history. Asqav signs each record server-side with keys held by an operator unaffiliated with the agent's operator, which supports the recordkeeping obligations the Exchange lists.
The recognition also reaches the verification protocol. An authoritative Asqav receipt is reproducible: the server re-derives the subject it signs, and any third party can re-derive the same record and check the ML-DSA-65 signature against the public keys Asqav publishes, with no callback to Asqav. A reviewer does not have to take Asqav's word for it, which is the property a tamper-evident log needs.
Read it
- The citing page: OWASP AI Exchange, input threats, secure logging.
- The profile: draft-marques-asqav-compliance-receipts on the IETF Datatracker.