Data processing addendum · effective September 19, 2026

How we process personal data on your behalf.

This addendum is part of the terms of service between you and Iron Rod Systems LLC, a Utah limited liability company doing business as Iron Rod AI™ ("we", "us"). It applies wherever you submit personal data to TruthLock and data protection law makes you its controller and us your processor. It takes effect when you accept the terms. Nothing needs signing, and we will sign a counterpart if your procurement process wants one.

1. Which data this covers

Two kinds of data move through TruthLock, and they are not governed the same way.

  • Customer content is what you send to be verified: responses, questions, citations, evidence retrieved from the connectors you declare, and the standing records and attestations produced from them. You decide what to send and why. You are the controller. We are your processor, and this addendum governs.
  • Account data is who your people are, how to reach them, and how you pay. We decide how to handle it in order to run a business, so we are the controller of it and our privacy policy governs instead. This addendum does not apply to it.

Where a term is defined in the GDPR, the UK GDPR, or the CCPA, it carries that meaning here.

2. What we do with it, and why

The full description sits in Annex I. In short: we process customer content only to run verifications you asked for, to keep the service secure and available, and to support you when you ask. We process it for as long as your account or stack exists, subject to the retention you set per organization, and then as section 10 says.

We do not sell it, share it for advertising, use it to serve another customer, use it to build or improve our own models, or mine it for statistics. Our own product metrics count verifications, latency, errors, and spend. They do not look inside what was verified.

3. Your instructions

We process customer content only on your documented instructions. The terms, this addendum, the configuration you set in the application, and the API calls you make are those instructions, and they are complete. If you want something else, write to legal@truthlock.io and we will agree it in writing, including who pays if it costs something to build.

We will tell you if an instruction appears to break data protection law, and we may pause that processing while we sort it out. If a law we are subject to compels us to process beyond your instructions, we will tell you before doing so unless that same law forbids telling you.

4. Who can see it

Access is limited to the people who need it to run or support the service, each under a written confidentiality obligation that survives their leaving. On a pooled deployment, access is scoped, logged in the audit trail, and available to you. On a dedicated stack, the data lives in an AWS account under your control and we have no standing path into it; we see what you show us in a support session, and nothing else.

5. Security

We keep the technical and organizational measures in Annex II, judged against the risk to the people whose data this is. We may change a measure as the product changes, but not in a way that materially weakens overall security. The security page carries the working detail and stays current.

6. Subprocessors

You give us general authorization to use subprocessors. Those that may touch customer content are named in Annex III. Each is bound by written terms no less protective than these, and we remain responsible to you for what they do.

Before adding or replacing one, we will give you at least 30 days' notice by email to your account owners. If you have a reasonable data protection objection, tell us within that window and we will work to offer you an alternative. If we cannot, you may terminate the affected part of the service and we refund prepaid fees for the unused period. That is the remedy.

7. When someone asks about their data

People whose data you sent us deal with you, not with us, because you are the controller. If one of them comes to us directly, we will not respond on the substance; we will tell them to contact you, and tell you promptly.

The application is built so you can answer them yourself: records are searchable, exportable, and deletable, and retention is configurable per organization. Where the tooling is not enough, we will help, and we will not charge for reasonable assistance.

8. If there is a breach

If we become aware of a personal data breach affecting customer content, we notify you without undue delay, and we aim to do it within 72 hours. The notice describes what we know: the nature of the breach, the categories and rough number of records, the likely consequences, what we are doing about it, and who to talk to. We will send what we know first and fill in the rest as we learn it, rather than wait for a complete picture.

We do not notify regulators or affected people on your behalf unless you ask us to in writing, because those are the controller's calls. We will not make a public statement identifying you without your consent unless the law compels it.

9. Impact assessments

If you need to run a data protection impact assessment or consult a supervisory authority about your use of TruthLock, we will give you the information we hold that you reasonably need and cannot get elsewhere.

10. Getting it back, and deletion

You can export your records at any time while the account is open, as a bundle that verifies on its own.

After termination you have 30 days to export. Then we delete customer content from live systems, and backups containing it roll off within 35 days. We keep nothing except what a law requires us to keep, which we then isolate and stop processing for any other purpose. A dedicated stack is in your own AWS account and is yours to keep or destroy; deletion there is your action, not ours. We will certify deletion in writing if you ask.

One honest limit. An attestation signs a hash of a record, not the record itself. Deleting the underlying content removes the readable data, and the hash that remains cannot be turned back into it. If you enabled the optional public-chain anchor, that hash is on a public ledger permanently and neither of us can remove it. It contains no personal data, but it does not disappear, and you should know that before you turn the anchor on.

11. Showing our work

We will give you the information you reasonably need to show that we are meeting this addendum, including our current security documentation and any third-party reports we hold.

If that is not enough, you may audit us, or have an independent auditor who is not our competitor do it, once in any twelve months and on 30 days' notice, during business hours, without disrupting the service, and under confidentiality. A regulator with the authority to audit may do so whenever its rules say. You cover the cost unless the audit finds we were in material breach, in which case we do.

12. Sending data across borders

Our infrastructure is in the United States, so using the pooled service means customer content is processed there. A dedicated stack runs in the region you choose.

Where you transfer personal data from the EEA to us, the European Commission's standard contractual clauses of 4 June 2021 apply, module two, controller to processor, and are incorporated here in full. Annex I of this addendum populates their Annex I, Annex II populates their Annex II, and Annex III their list of subprocessors. Clause 7 does not apply, clause 9 uses option two with the 30 days' notice in section 6, clause 11 does not use the optional independent dispute body, clauses 17 and 18 take the law and courts of Ireland, and clause 13 takes the supervisory authority of the member state in which you are established or represented.

For the UK, the Information Commissioner's international data transfer addendum applies to those clauses, with tables 1 to 3 completed by the annexes here and neither party permitted to end it under section 19 except as that addendum requires. For Switzerland, the clauses read with the Federal Data Protection Act substituted for the GDPR and the Federal Data Protection and Information Commissioner as the authority. If any of these instruments is replaced, the replacement applies from the date it takes effect.

We have no establishment in the EEA or the UK. Where the law requires a representative for the processing we do here, we will appoint one and tell you who it is.

13. California

For personal information covered by the CCPA, we are a service provider and you are the business. We do not sell or share it as that statute defines those words. We do not retain, use, or disclose it for anything outside the direct business purpose of providing TruthLock to you, and never for a commercial purpose of our own. We do not combine it with personal information from any other source, except as the statute permits a service provider to do. We understand these restrictions and we comply with them. You may take reasonable steps to confirm that we do, which section 11 describes.

14. Precedence, liability, and law

Where this addendum and the terms of service conflict on the handling of personal data, this addendum wins. On everything else, the terms win. Each party's liability under this addendum is subject to the limits in section 13 of the terms, counted together with the terms rather than in addition to them, except where data protection law forbids limiting it.

This addendum is governed by the law of the State of Utah and disputes follow section 18 of the terms, except where the standard contractual clauses require otherwise, in which case they govern themselves. If a provision is held unenforceable, it is narrowed to the least extent needed and the rest stands.

15. Putting it in place

Accepting the terms of service accepts this addendum, so there is nothing you need to do. If your process requires a signed counterpart naming your legal entity, write to legal@truthlock.io and we will send one that matches this text.

Annex I · The processing

Parties

Data exporter and controller: the customer named on the account, whose contact details are the account and billing contacts on its company record. Its role is controller of the customer content it submits.

Data importer and processor: Iron Rod Systems LLC, dba Iron Rod AI™, Utah, United States. Contact privacy@truthlock.io and 833-IRON-ROD. Its role is processor, providing AI answer verification.

Categories of people

Whoever appears in the material you submit. Typically that means your own customers, employees, members, or the subjects of the answers you verify. You choose, and you are the only one who knows in advance.

Categories of personal data

  • Whatever appears inside the AI responses, questions, and citations you submit.
  • Whatever appears inside the evidence retrieved from the connectors you declare.
  • The claims, grades, reason codes, and standing records derived from those.
  • Identifiers your systems attach to a request, such as an end-user reference you pass through.

Special category data

TruthLock is not designed for it and we ask you not to send it. If your use case requires it, tell us first at legal@truthlock.io so we can agree the extra restrictions in writing rather than discover them later.

How often, and how long

Continuously, for as long as you use the service. Retention follows the setting you configure per organization, and otherwise section 10.

Nature and purpose

Decomposing responses into claims, retrieving and snapshotting evidence, grading each claim against the source hierarchy you declared, applying your reliance policy, producing signed standing records and attestations, and storing and serving all of it back to you. Plus the ordinary work of hosting: backup, monitoring, security, and support.

Subprocessor processing

As Annex III describes, for as long as the subprocessor's role requires.

Competent supervisory authority

For the standard contractual clauses, the authority of the EEA member state where you are established, or where your Article 27 representative is, or where the affected people are.

Annex II · Technical and organizational measures

  • Encryption. TLS in transit. Encryption at rest for the database, object storage, and backups, under keys held in AWS KMS. Signing keys are asymmetric and their private half never leaves KMS.
  • Tenant separation. A dedicated deployment is its own VPC in its own AWS account with no shared control plane. On the pooled tier every table is scoped by company, every route is tested for scoping, and Postgres row-level security enforces it beneath the application, so a bug in the application layer is not enough to cross the boundary.
  • Authentication. Argon2id password hashing, optional authenticator app, server-side sessions behind an httpOnly cookie, throttled failed attempts, sign-out everywhere, or OIDC to your own identity provider.
  • Authorization. Role-based, with one capability table deciding what the interface shows and what every API route refuses, so the API cannot do what the screen would not allow. Nobody can grant a capability they do not hold. API keys are hashed at rest and scoped to an organization and a capability set.
  • Secrets. Connector credentials and any model provider key you paste are encrypted in AWS Secrets Manager in a space reserved for your company, shown masked, never returned in full, and rotated by replacement.
  • Integrity. Standing records are serialized canonically, hashed claim by claim into a Merkle tree, and the root is signed. Attestations are append-only and enforced as such in the database: a correction issues a new attestation linked to the old one, and the old one keeps validating while stating that it was superseded and why. Nobody, including us, can quietly alter a record.
  • Logging. An audit log of sign-ins and failures, invites, role and team changes, key and organization changes, policy edits, and adjudications, each with actor, before, and after. Application logs and traces go to CloudWatch inside the stack.
  • Availability. Managed AWS services with automated backup, infrastructure defined in Terraform and rebuildable from source, and health-checked rolling deployments.
  • Minimization. Redaction hooks run before storage when you configure them. Retention is set per organization. Evidence is stored as a hashed snapshot with its fetch time rather than re-fetched later.
  • Provider boundaries. Every record names the provider, the model, and the last four characters of the key that produced the grade. Bring your own key and the calls run under your provider agreement, so any zero-retention terms you hold apply.
  • People. Access limited to those who need it, under confidentiality obligations that survive their departure, removed when they leave.
  • Change control. Source control with review, automated tests including tests that assert tenant scoping, and migrations applied as a tracked step of deployment.

Annex III · Subprocessors

These may process customer content.

  • Amazon Web Services, United States, us-east-1 by default or the region of your dedicated stack. Compute, database, object storage, key management, and secrets.
  • Model providers, United States. Anthropic by default, and OpenAI, Google, xAI, or Amazon Bedrock where an organization selects them. They receive the material being verified and the evidence retrieved, in order to extract and grade claims. Under our agreements they are prohibited from training on it. Bring your own key and your agreement governs instead.
  • Exa, United States. Live web search, only where an organization enables the website connector, and only across the domains you declare. It receives the search terms derived from what you submitted.

These process account and support data, where we are the controller and the privacy policy governs rather than this addendum. They are listed for completeness, not as subprocessors of customer content.

  • Cloudflare for this website, its rate limiting, and delivery of contact-form submissions.
  • Fastmail for the mailboxes behind our published addresses.
  • Stripe for payments, invoices, and receipts.
  • Google Fonts, which your browser contacts directly to load typefaces on this website.