Audit AI agents the way your risk committee actually wants:
cryptographically, continuously, vendor-independently.
Vortalis records each AI agent action routed through the service proxy on a tamper-evident, hash-chained audit log, issues identity attestation tokens signed with ES256, tags audit rows with the regulations that apply, and exports signed evidence packs your auditor can verify in their own tooling. No callbacks and no proprietary viewers; verification uses an offline utility and our public key set.
Three audit layers, one platform
Audit isn't one job. Vortalis serves all three lines of defence and the external audit firm at the same time, from the same data.
Continuous policy enforcement at the agent layer. Every action routed through the service proxy is decided, hash-chained into the audit log, and recorded in tens of milliseconds. Your operators see compliance as a property of the platform, not a separate workstream.
Real-time, machine-readable compliance reports for the EU AI Act, DORA, MiFID II, and the other frameworks that have an evidence pack builder, without manual evidence collection. FCA Consumer Duty is an audit tag only, with no pack builder. Findings, remediation, and sign-off built into the audit flow.
Signed evidence packs the auditor consumes in their own tooling. Offline verification against a saved copy of the public JWKS. Same data, same answer; once the auditor holds the public keys, verification does not call our servers. Maps to ISO 42001, NIST AI RMF, and DORA Article 28.
What's in an evidence pack
The artefact your auditor actually wants, generated from the audit log on request.
Pack metadata: tenant, period, filters, counts, Merkle root, regulatory mapping with regime versions declared at export time.
Detached ES256 signature over canonicalised manifest bytes. Verifiable against the public JWKS.
Each recorded agent action (allowed, denied, or held for approval), with chained entry hashes for tamper detection.
Identity Attestation Tokens: signed JWTs proving who the agent was at the time of each action.
Verifiable Intent Artefacts: signed JWTs proving what the agent was authorised to do, per major card-network agent-pay protocols.
Snapshot of active and rotated public keys at export time. Lets the verifier reproduce the cryptographic state.
Each regulatory tag with the regime version declared at export time: EU AI Act, DORA, MiFID II, and others.
Per-entry previous_hash → entry_hash pairs for independent O(n) chain re-validation.
Six-line Python verification snippet so the auditor verifies in their own tooling without help.
Regulatory coverage on day one
Every audit row carries tags for the regimes that apply to the service it touched. The regime version is declared at export time, not stamped when the action happened. No mapping spreadsheets, no compliance officer tagging.
The first group below has an evidence pack builder in the compliance engine. The second group is carried as an audit-row tag only and has no pack builder. The DORA mapping shown here, and the CSDR/SDR mapping on our financial-services page, specifically carry explicit internal caveats and await external counsel review before we would rely on them, unassisted, in a live regulatory submission; treat those two as engineering-ready rather than legally sworn off.
Evidence pack builders
Audit tags only (no evidence pack builder)
Ask for a signed evidence pack and verify it yourself.
A sample pack is available on request. Its manifest signature can be checked offline against the public JWKS. There is no public in-browser verifier on this site today.