Your security documents deserve better than a promise

Rendro is an early-stage company. We keep the system small, build on independently audited infrastructure providers, and switch on strong controls by default rather than inventing our own.

You upload your own documents and a questionnaire. For each question, Rendro finds the relevant passages in your documents, drafts an answer that cites them, and flags anything it can't support. You review and approve every answer before exporting.

• Application: Vercel (serverless, United States).
• Database, sign-in and file storage: Supabase (PostgreSQL), hosted on AWS in us-east-1 (United States).
• During processing, text is sent to our AI providers in the United States (see AI and your data).
• We run no servers of our own. The full list is on our subprocessors page.

• In transit: every connection uses HTTPS (TLS), and browsers are told to always use HTTPS for rendro.net (HSTS). Connections from Rendro to its providers use TLS too.
• At rest: the database and stored files are encrypted at rest by our database provider, Supabase.
• Secrets: keys live in encrypted environment settings, never in source code. The database's administrative key is used only on the server.

• Every table holding customer data carries a workspace id, and PostgreSQL Row Level Security is on for every one of them: the database itself refuses to return another workspace's rows, even if application code had a bug.
• Automated tests sign in as one workspace and check they can't read, list or change another's data or files.
• Files sit in private storage under a per-workspace folder. Downloads use signed links that expire after 60 seconds.
• AI search only looks at the requesting workspace's documents. One customer's content is never used to answer another's questions.

• Sign-in with email and password (12+ characters, email confirmation required), with rate limits against guessing.
• Roles per workspace: Owner, Admin, Editor, Reviewer. Reviewers can comment and suggest, but can't approve, export or change settings.
• Sensitive actions (invitations, role changes, uploads, deletions, exports) are recorded in an audit log.
• Two-factor sign-in for customer accounts and single sign-on (SAML) are not available yet; both are on our roadmap.
• Inside Rendro, production access is limited to the founder. We don't look at customer content except when you ask for support or to investigate a security or reliability problem.

• Every input is validated on the server; uploads are limited to 25 MB and to allowed file types.
• Uploaded content is treated as untrusted: hidden characters are removed, and the AI is told to treat documents and questions as data and ignore any instructions hidden in them (prompt-injection defences, checked by tests).
• Exported answers can't turn into spreadsheet formulas, and we write only into your answer cells (formatting, dropdowns and formulas are left untouched).
• Browser protections on every page: HTTPS only, no framing by other sites, no content-type guessing, and a strict Content Security Policy so pages only run Rendro's own scripts.
• Rate limits on sign-in, uploads and AI use.
• Daily spending limits on AI use for every workspace.
• Private source code. Automated checks (type check, lint, tests and build) are part of how we make changes, and dependency alerts are on.

• We don't train AI models on your data. We don't train models at all; we use AI providers through their business APIs.
• Anthropic (Claude) drafts answers. Under Anthropic's commercial terms, data sent through its API isn't used to train its models by default.
• Voyage AI turns the text of your documents and Answer Library, in passages, into search vectors (embeddings) so the right passage can be found, and does the same for each question.
• Only what each step needs is sent. To draft an answer, Anthropic receives the question, your company profile and the few most relevant passages from your workspace, not your whole library. To find a questionnaire's question and answer columns, it sees a short preview of each sheet. If you use "prefill from website", it sees the public text of your website's home and /security pages.
• Every answer shows its source and a confidence level. Answers your documents don't support are marked "Needs review" instead of guessed, and nothing is exported without your review.

• Your data is kept while you use Rendro.
• Deleting a document removes the file, its extracted text and its search vectors. Deleting a questionnaire removes the file, its questions, answers and exports, immediately.
• On request, we delete all of a workspace's data within 30 days.

If we become aware of a breach affecting your data, we'll notify you without undue delay and within 72 hours after becoming aware, with what we know and what we're doing, and report to regulators where the law requires it (including PIPEDA in Canada).

We'd rather be honest than impressive:

• SOC 2 Type I / II: not started. Planned as the business grows or when customers need it.
• Independent penetration test: not done yet.
• Our infrastructure providers hold their own audits (for example SOC 2). That doesn't make Rendro SOC 2 compliant, and we don't claim it does.

Found a security issue? Email security@rendro.net. We'll acknowledge it within 3 business days. Please give us reasonable time to fix it before sharing it.

Need our own security questionnaire answered? Ask us, and we'll send it back filled in with Rendro, of course.