AI Disclaimer

AI-Assisted Output

Mazora reports and guardrail work item drafts are generated with AI assistance, using OpenAI, from your saved blueprint context. Outputs may be incomplete, inaccurate, outdated, or unsuitable for your specific product.

Suggested Context Wording

When a work item mentions something your context does not describe, Mazora can suggest how the relevant context section should read with that fact included. Noticing the gap is deterministic and involves no AI; the suggested wording is AI-generated and carries every caveat on this page.

Two things are worth knowing before you accept one. The suggestion is written only from the sentences in that work item, so anything those sentences leave unsaid can be filled in wrongly or stated with more confidence than the work item earned. And accepting replaces the whole section, not just the new sentence — Mazora checks whether the suggestion dropped anything you had written and lists what it can find, but that check compares wording and can miss a meaning that was quietly changed rather than removed.

Read the suggestion before accepting it, and edit it where it overreaches. Accepted text becomes ordinary context, which is what Mazora sends to the AI provider and what briefs your coding agent.

Scanned Codebase Context

If you connect a repository, Mazora works out your tech stack by matching dependency manifests against a fixed list of known packages. That matching involves no AI, but it is still a guess: it can miss things you use and report things you no longer use. A dependency listed in a manifest is not evidence that it is actually in use.

Those findings are then written up by OpenAI, so the summary you read is AI-generated and carries the same caveats as any other AI output here — it can overstate, omit, or misdescribe what the matching actually found.

Scanning Actors, Policies, or Workflows goes further: Mazora selects application source files by name and asks OpenAI to describe what they show. Two things follow. The selection is a guess — it can read the wrong files, or miss the ones that matter. And a description is only ever drawn from the files that were read, never the whole repository, so treat "the code does not appear to do X" as a statement about a sample rather than about your product.

Treat a scanned tech stack as a first draft to correct, not a finding to rely on. It is saved as ordinary editable context, and anything left uncorrected becomes part of what Mazora sends to the AI provider when you generate a report.

Review Before Use

Review every report before using it with developers, coding agents, investors, customers, or internal teams. Treat Mazora as a planning assistant, not a final authority.

Not Professional Advice

Mazora does not provide legal, financial, security, compliance, medical, or tax advice. Consult qualified professionals for high-stakes decisions.

Sensitive Information

Do not paste secrets, credentials, access tokens, private keys, regulated personal data, or confidential information that should not be sent to AI providers.

This applies to scanned context too. Review what a scan wrote before generating a report, and remove anything that should not leave your systems.