Privacy Policy

Last updated 13 August 2026 · KRAYON-AI, LLC · Florida

This policy is written to be read, not to be survived. It covers three different kinds of data, and the rules are genuinely different for each: what our website collects, what our conversational products handle, and what happens to client documents. If you only care about the third one, jump to section 3.

1. Who we are

KRAYON-AI, LLC is a Florida limited liability company (document number L26000410768) operating from Tampa, Florida. In this policy, "we" and "Krayon" mean KRAYON-AI, LLC. "You" means either a visitor to this website or a client who has engaged us for services, depending on context.

For most client engagements involving documents, we act as a service provider or processor: you decide what we process and why, and we process it only on your instructions. Where the data is protected health information, we act as a business associate under HIPAA once a Business Associate Agreement is in place. See section 9.

2. Data the website collects

This site is deliberately light. There is no advertising network, no third-party analytics script, no session recorder, no cross-site tracking pixel, and no cookie banner because we do not set tracking cookies.

• Page view counts. A page view is counted server-side with no cookie, no device fingerprint, and no identifier tied to you. We can tell that the staffing page was viewed; we cannot tell that it was you.
• Standard request logs. Our hosting provider processes ordinary web request data, including IP address, in order to serve the page and to block abuse. These are retained on a short rolling window and are not used to build a profile of you.
• What you type into a form. If you submit the assessment, contact, or BAA request forms, we receive what you typed and your email address. We use it to reply to you and to keep a record of the request. Nothing more.
• Rate limiting. Interactive endpoints keep a short-lived, per-IP counter to prevent abuse. It expires automatically and is not linked to any other record.

Local browser storage is used on some pages to remember interface preferences, such as a dismissed banner or a conversation you have had with the public demo. That data stays in your browser and is never transmitted to us. Clearing your browser data removes it.

3. Client documents and the air-gapped pipeline

This is the section that matters if you are a law firm, an insurance carrier, a medical billing company, or anyone else whose files legally cannot leave the building.

Documents you send us for processing are processed on computers we own and physically control in Florida. They are not uploaded to OpenAI, Anthropic, Google, Microsoft, Amazon, or any other third-party artificial intelligence service. There is no public-cloud model API anywhere in the document pipeline. The models that read your files run on our own hardware, and the machines doing that work do not send your file contents anywhere else.

Isolation. Privileged document work is separated per client. Where a client requires it, we assign dedicated hardware so that one client's files never share a machine or a disk with another client's files. This is available on request and is the default for engagements involving protected health information.

Human review. Work product is reviewed by a person before it is delivered to you. That reviewer is either the principal of this company or a contracted reviewer bound by a written confidentiality agreement. We do not ship unreviewed model output as a finished work product.

Transmission. Files are transmitted over encrypted connections. We will accept delivery through your own secure portal or file-transfer system if you prefer to keep the chain of custody inside tooling you already trust.

4. Conversational products (SMS and chat)

Our conversational products — the AI sales employee, the staffing employee, and the public demo — work differently from the document pipeline, and we would rather be precise about it than let a general privacy claim cover something it does not cover.

If you leave a cell number or email after using the public demo, we keep only that contact detail, an optional name, and which demo lane you were in — never the demo conversation itself, which is not stored — and we use it once, by a person, to reach you about your own number.

These products may use commercial language-model providers to generate replies. Message content in a conversational engagement can therefore be transmitted to a model provider under a commercial agreement that prohibits training on that content. This is not the air-gapped pipeline described in section 3, and we do not represent it as such.

If you need a conversational deployment that also runs entirely on our own hardware, that is something we can build, and it is priced differently. Ask on the call rather than assuming either way.

For text-message deployments we also handle: opt-out keywords and honoring them across the whole system, quiet hours, and message frequency caps. Opt-out records are retained deliberately, because deleting them would cause someone who asked not to be contacted to be contacted again. Retaining a suppression record is a privacy protection, not a violation of one.

Mobile numbers and consent records are never sold or shared for marketing. No mobile information will be shared with third parties or affiliates for marketing or promotional purposes. Text-messaging originator opt-in data and consent are not shared with any third party — the only exceptions are the carriers and messaging providers that physically deliver the message, as described in section 8. In any text conversation with us or with a business we operate messaging for: reply STOP to opt out at any time, reply HELP for help. Message frequency varies by conversation. Message and data rates may apply.

5. We do not train on client data

We do not use client documents, client message content, or any data you give us to train, fine-tune, or otherwise improve models for any other client or for our own general-purpose products. Where a commercial model provider is used as described in section 4, we use configurations that contractually prohibit training on submitted content.

If we ever want to use something you gave us to improve a system beyond your own engagement, we will ask you in writing first, and "no" is a complete answer that costs you nothing.

6. Retention and deletion schedule

Published schedules are worth more than promises, so here is ours. All periods run from the date we deliver your work product.

• Source documents you sent us: deleted after 30 days. This is the pile of records, the claim file, the export. It is removed from working storage and from any backup image on the schedule below.
• Delivered work product: deleted after 90 days. We keep the chronology or summary for a short window so that we can re-send it or answer a question about it. After that it is gone and you are the only holder.
• Access logs: retained 12 months. Logs record who touched a file and when. They are retained longer than the files themselves on purpose — a deletion log that gets deleted proves nothing.
• Business records: invoices, engagement letters, and correspondence are retained as long as required for tax and legal purposes.
• Backups: any backup containing client documents is on a rotation not longer than the retention period for the data it holds, so a deleted file does not survive in a backup indefinitely.

Deletion on request. You can ask us to delete your data earlier than the schedule at any time, for any reason or none. We complete the deletion within five business days and confirm it in writing. There is no charge and no exit interview.

Longer retention on request. Some firms need the opposite — a file held for the life of a matter. That is a written agreement, not a default, and it is priced into the engagement.

7. Access control and logging

Access to client documents is limited to the people who need it to do your job. Access to your files is logged for the life of your matter with the identity of the accessor, the file, and the time, and we provide that log to you with the work product rather than only on request. To be exact about what it is: a per-engagement record maintained by a small firm, not an enterprise audit platform. If your requirements call for tamper-evident or independently attested logging, raise it before engaging us.

Full technical detail is on our security page.

8. Subprocessors

We keep this list short on purpose, and it is worth understanding the boundary: none of the providers below have access to client documents. They support the website and business communication only.

• Cloudflare — hosting, content delivery, and abuse protection for this website, plus the storage behind the website's forms.
• Resend — transactional email delivery, so that a form submission reaches us and you get a confirmation.
• Telecommunications carriers and messaging providers — for text-message products only, to actually deliver a text message. This is unavoidable; text messages travel over carrier networks.
• Commercial language-model providers — for conversational products only, as described in section 4, under agreements that prohibit training on submitted content.
• Contracted reviewers — where a licensed reviewer is used for document work product, bound by a written confidentiality agreement.

We will give you the current list in writing on request, and we will tell you before adding a subprocessor that would touch your data.

9. HIPAA and protected health information

Medical chronology work routinely involves protected health information. Where you are a covered entity or a business associate and you are disclosing PHI to us, we sign a Business Associate Agreement before any protected health information is transmitted. You can request one here, and you can have it in hand before a sales call, not after.

Under a BAA we use and disclose PHI only as the agreement permits, apply appropriate safeguards, report security incidents and breaches to you as required, bind any subcontractor to the same terms, and return or destroy PHI at the end of the engagement.

Personal injury firms should note that records obtained in litigation are frequently not covered by HIPAA in your hands, but are still privileged and still confidential. We treat them to the same standard either way.

10. Your rights and how to exercise them

Depending on where you live, you may have the right to know what personal data we hold about you, to get a copy of it, to correct it, to have it deleted, and to opt out of its sale or sharing. We do not sell personal data and we do not share it for cross-context behavioral advertising. There is no dark pattern here to route around: email king@krayon-ai.com with the words "privacy request" and describe what you want. We respond within 30 days and will not charge you or make you create an account to ask.

If you are an individual whose information appears inside a client's documents rather than a client of ours, we process that data on our client's instructions. Send your request to us and we will route it to the client responsible for it, and tell you that we have done so.

11. Incidents

If we discover a security incident affecting your data, we notify you without unreasonable delay and within any period required by law or by our agreement with you, and we tell you what we actually know rather than waiting until the story is tidy. Our incident response approach is described on the security page.

12. Changes to this policy

If we change this policy in a way that materially reduces the protections described here, we will notify active clients directly rather than quietly updating the date at the top. The last-updated date always reflects the current version.

13. Contact

Privacy questions, deletion requests, subprocessor lists, and access logs:
king@krayon-ai.com
813-733-5997
KRAYON-AI, LLC · Tampa, Florida

This policy describes our practices. It is not legal advice to you, and it does not replace the terms of a signed agreement between us — where a signed agreement and this page differ, the signed agreement governs.