Prove what your AI acted on.
When an automated decision (a claim denial, a credit or pricing call, an eligibility decision) is later contested or subpoenaed, the question is what external data the system acted on, at that moment. Dynamic Feed is the neutral third party that signs and timestamps that input datapoint, independent of the system that decided, so it holds up. We are not the decision-maker and we do not supply the data. We witness the external input.
re-check against /.well-known/keys
When discovery asks what the model saw.
In US litigation, courts are increasingly ordering disclosure of the data behind an automated decision: not just the model and its output, but the external inputs the system acted on when it denied a claim, priced a policy, or scored an applicant. The order log and a model card do not answer that question on their own.
Saying "the system used a third-party feed" is not independent, and a value reconstructed later under discovery can be argued. Dynamic Feed is the independent witness of the external data feeding the decision. Because Dynamic Feed is not the deciding party, the signed record is independent by construction, and independence from the deciding party is the value when the other side reads every input as self-serving.
ICO contest rights, reconstructed.
Under UK data-protection rules, a person subject to a significant automated decision has the right to contest it, and the ICO's expectation is that a firm can reconstruct which system ran the decision and what data sources it drew on. The second half is the hard half: showing, after the fact, the exact external datapoint the system consumed at decision time.
Dynamic Feed gives you the signed record of the external data the system drew on: captured contemporaneously, timestamped, and re-checkable by you or the regulator. It is one independent, verifiable input toward a contest response, captured as it happened rather than reassembled when the request lands.
A signed receipt for every decision's inputs.
You keep consuming your own external data. For each automated decision, you hand us the input it acted on, or just its digest, and we hand back evidence.
- Sign the input, or only its hash. Send the value or a SHA-256 digest. Send only the digest and the underlying data never leaves your systems, so there is nothing for us to read or store.
- Ed25519-signed receipt. Every receipt carries the source and exact measurement timestamp of the external input, signed over its precise bytes.
- Bitcoin-anchored proof-of-existence. Anchor the receipt's fingerprint to the Bitcoin blockchain, independent evidence it existed, unchanged, no later than a given block.
- Append-only archive. Records are written as they happen to a permanent store, so they cannot be quietly rewritten before discovery or a contest.
- Independent re-verification. You, a court, or a regulator re-check any receipt against the published key, with no trust in Dynamic Feed required.
- Neutral by design. Dynamic Feed never makes or reviews the decision and never supplies the data. It witnesses the external input, nothing else.
Don't take our word, verify a live record.
This fetches a live signed record from Dynamic Feed and verifies its Ed25519 signature in your browser, against the published key. Then flip one byte and watch verification fail, the same tamper-evidence a court or regulator would rely on, demonstrated live.
One keyless call, a signed receipt.
Notarise the input
POST the external datapoint your system acted on, or just its SHA-256 digest, to a keyless endpoint.
Get a signed receipt
Back comes an Ed25519-signed, optionally Bitcoin-anchored receipt with source and exact timestamp.
Keep it against the decision
Store the receipt against the decision id. It is written to the append-only archive, so you can produce it later.
Anyone re-verifies
You, a court or a regulator re-check the signature on the public verify page or with the open-source verifier.
Teams who must show their working.
Input evidence, on file
For every automated claim or pricing decision, an independent signed record of the exact external datapoint it acted on, retained and re-checkable if the decision is later contested or subpoenaed.
Defensible decision inputs
Show precisely which external data the model consumed at decision time, captured as it happened, not reconstructed once an applicant disputes the outcome.
A record you can reconstruct
When an automated eligibility decision is challenged, produce the signed external input the system drew on, re-verifiable without trusting your own logs.
A signing seam to embed
A keyless MCP and REST endpoint your platform can call to attach independent evidence to the decision records you already capture.
Evidence that re-checks
A cryptographic, third-party record that an external input existed unchanged at a time, useful when a decision is later questioned in discovery or an ICO contest.
The witness your system cannot be
An automated system cannot witness itself. Dynamic Feed gives you and your counsel an independent record of the external data behind each automated decision.
Evidence, not a certification.
Dynamic Feed is an independent evidence layer, not the decision-maker, not a party to the decision, not legal advice, and not a compliance certification. It does not supply the data the system used and does not make or review the decision. It signs the external input datapoint your system already consumed, as an independent witness. Signing and anchoring prove a datapoint existed, unchanged, at a specific time, and is independent of the deciding party; they do not prove the value was correct or that the decision was right. This is tamper-evident, not tamper-proof. Treat it as one independent, verifiable input toward your own record of what data a decision drew on, never the sole basis for a legal or regulatory outcome.
Fully automated. Run by AI, end to end.
No sales calls and no waiting. Mint a key and start notarising your decision input evidence in seconds.