AI & Data Use

1. Summary

SyndTrack uses AI for three things: parsing uploaded documents, parsing uploaded images, and scoring a deal you have entered. Each sends a specific, limited set of information, only when you choose it, and each has a way to accomplish the same task without AI. CSV import uses no AI at all. Nothing is sent to an AI provider in the background. The full policy below controls.

2. Current AI provider

SyndTrack currently uses Anthropic's direct commercial Messages API. SyndTrack does not currently use the Files API, Message Batches API, prompt caching, or beta APIs.

• Anthropic — direct commercial Messages API — retention terms

3. Document parsing

When you choose to parse a PDF or text document, SyndTrack extracts text from it and sends up to 100,000 characters of that text to Anthropic to identify portfolio fields. The original file is not sent through the provider's persistent file-storage API. Extracted text can include names, addresses, tax identifiers, account details, or wire instructions contained in the document.

4. Image parsing

Images are different from documents, and the difference matters: the complete image is encoded and sent to Anthropic. Anything visible in it is included, including information that text extraction would never have picked up. Redact full tax identifiers, bank-account numbers, and wire instructions before using AI parsing on an image.

5. CSV mapping

No part of a CSV import is sent to an AI provider. Column mapping is done deterministically on our own servers by matching your headers against a table of known aliases. Your spreadsheet, its headers, and its rows stay within SyndTrack.

An AI-assisted mapping endpoint existed in the codebase from April 2026 but was never connected to the product and never received a request from it; it was deleted in August 2026. We mention it because "we could have sent it and did not" is a more useful thing for you to know than silence.

6. Deal scoring

AI scoring sends the structured deal and sponsor information shown on the deal record to Anthropic. It does not send the contents of uploaded documents. The score is based only on what you entered, does not use live market data, may be incomplete or inaccurate, and is not an investment recommendation.

7. What is not sent, in each workflow

• Document parsing: the original file itself is not sent to a persistent file-storage API, and text beyond the disclosed character limit is truncated before transmission.
• Image parsing: no other document, and no portfolio record, accompanies the image.
• CSV import: nothing at all — no headers, no rows, no file. Mapping never leaves our servers.
• Deal scoring: no document contents, no other deals, and no market data.
• Across all four: no billing details, no authentication credentials, and no data belonging to another account.

8. Retention by SyndTrack and current AI providers

SyndTrack stores the document, the extracted text, and the parsed fields in your account until you delete them. Anthropic states that standard commercial API inputs and outputs are deleted from its backend within 30 days, except where a different agreement applies or longer retention is needed for legal or usage-policy enforcement. Their current terms are linked in section 2 and are theirs to change.

9. Model training and feedback

Anthropic states that commercial API data is not used to train its models by default; customer opt-in or deliberate feedback may create an exception. SyndTrack does not submit customer content as feedback through the application. If SyndTrack changes or adds a provider, that provider's applicable training and data-use terms will be disclosed here.

We deliberately do not yet make the stronger statement that our organization has opted out of training, because that claim requires provider-account evidence to back it. It will be added here once that evidence exists, not before.

10. Processing locations

SyndTrack's primary production database is hosted in a U.S. region. The application and content-delivery network operate globally, and subprocessors, including current or future AI providers, may process information in the United States and other regions under their contractual and security commitments. SyndTrack does not promise that all processing occurs exclusively in the United States.

11. Accuracy and human review

Every AI output is review-gated: parsed fields and scores are shown to you and do not change your portfolio records until you approve them. AI output can be incomplete, inaccurate, inconsistent, or outdated. Compare it against the source document before saving or acting on it.

12. Non-AI alternatives

Every workflow has one, and none of them is a degraded path: upload a document without AI parsing and enter the figures yourself, map CSV columns manually, and skip deal scoring entirely. Choosing them sends nothing to any AI provider.

13. Changes and contact

SyndTrack may add, replace, or discontinue AI models, model versions, and AI service providers as the Service evolves. Before your information is sent to an AI provider, the applicable feature notice will identify the provider or providers then in use. We keep this page and Security & Subprocessors current, and will give additional notice of a material change when required by applicable law or contract. A replacement provider must be subject to contractual privacy, confidentiality, and security obligations appropriate to the information it processes. Nothing here promises that any particular provider remains exclusive.

Questions: contact us. Methodology for scoring is documented at AI scoring methodology.