Notice: hum-beta-2026-11-04-v1 · Salient One (SalientOne OÜ).
Private service
Your verified Salient One identity owns each source package, file, report and follow-up. Organisation labels do not grant team access. General owner administration does not expose confidential proposals, document text, reports or private feedback explanations. Do not upload beneficiary or case-level personal data, secrets, or material you cannot authorise for processing.
Processors and limitations
Cloudflare stores private service records and converts PDFs using the existing Markdown conversion service. DOCX body text is parsed in the application. OpenAI processes the extracted text, supplied context and findings through its API. HUM requests no stored Responses history and does not opt into provider training or data sharing. This is not a promise of zero provider retention: OpenAI may keep abuse-monitoring content for up to 30 days under standard controls. Salient One must review the actual project’s provider settings and current credits before analysis is enabled.
Private pages have no third-party analytics scripts. Optional account-linked product events use the existing shared usage settings, consent revision, withdrawal and 90-day retention. Essential service, cost and abuse records operate separately and contain no raw document text.
Retention, saving and deletion
Unsaved packages, source text, original uploaded files, reports and follow-ups expire after 24 hours. An explicit private save extends the package for 90 days. Access is denied immediately at expiry; the existing cloud notification Worker removes expired content in bounded maintenance batches. Deleting a package or all beta data removes its content and private feedback and withdraws affected learning candidates and releases. Model processing already started may finish and count toward service limits. Essential cost metadata is retained for up to 90 days; aggregate budget totals have no proposal content. Infrastructure backups and processor retention follow their separate policies.
Optional shared improvement
With permission, checked analytical findings can generate candidates from a small fixed vocabulary of general issues, such as conflicting labelled dates or an unprovided required annex. The pipeline records permission revision and private provenance, then uses fixed synthetic cases without client wording or project identifiers. A person must separately validate the lesson, confidentiality boundary and synthetic test; representative, held-out and negative cases must pass before a versioned release can affect future analyses. Drafts, user edits and reported funding outcomes do not validate a claim. No AI approves its own claim as global knowledge.
Original proposals remain private. Automatic redaction is not a guarantee of anonymity. Identifying or confidential lessons must remain private and cannot be submitted as free-form shared examples. Permission withdrawal or source-case deletion withdraws affected releases; baseline rules remain available. Existing reports retain their dated method and knowledge version.
Optional research contact
This permits contact about product research and is separate from service use, shared improvement, call alerts and marketing. Declared unknowns and profile corrections are welcome. Future price interest is a reported preference, not a payment or automatic conversion.
Contact
engage@salientone.eu · Salient One privacy policy · OpenAI API data controls · Return to HUM PRO (Beta)