Your life is personal. ARDA treats it that way.
ARDA needs context to give you a useful outside view. That makes restraint, transparency, and control part of the product rather than fine print.
Only what the current task needs
ARDA is designed to send bounded context for one audit, explanation, plan, or conversation task rather than sending your full account by default.
Your records remain inspectable
Facts and evidence are designed to be reviewable, correctable, exportable, and deletable. Model assumptions do not silently become accepted evidence.
No advertising business model
ARDA does not sell private app content, build advertising audiences from it, or opt your content into general-purpose AI model training.
ARDA keeps the product record. AI handles a bounded request.
- 1You choose what to share
Audit answers, conversations, goals, and updates come directly from you. Sensitive questions can be declined where appropriate.
- 2ARDA stores the record in Supabase
Your account, accepted evidence, ratings, plans, and history live in ARDA’s application database. OpenAI is not ARDA’s long-term memory.
- 3Only relevant context is sent for the current AI task
ARDA’s server selects the minimum approved context, calls OpenAI through a secure API, validates the result, and keeps model output separate from accepted facts until the product flow confirms it.
What OpenAI receives, and what it does not.
For a feature that needs AI, ARDA may send the relevant audit answer, conversation excerpt, accepted evidence, or planning context. It does not send your email address, authentication secret, or full account by default.
Eligible requests are configured with response storage disabled through the API (store: false). ARDA does not opt API content into model training. Under standard OpenAI API controls, limited prompt and response content may still remain in abuse-monitoring logs for up to 30 days. ARDA will describe Zero Data Retention only if it is approved, enabled, and verified for production.
Private by default, including when you ask for help.
Support personnel do not receive default access to intimate app content. If content access is genuinely needed to resolve a request, the intended design requires it to be purpose-limited, time-limited, and audited.
ARDA is not a therapist, doctor, crisis line, financial adviser, or emergency monitor. Safety-related language does not automatically lower a rating and is never used to pressure someone into subscribing.
Read safety and support information- Abandoned anonymous audit
- Up to 7 days
- Temporary export file
- 24 hours
- Resolved support request
- Normally 90 days
- OpenAI abuse-monitoring logs
- Up to 30 days
The complete schedule and exceptions are described in the Privacy Policy.
The companies needed to operate ARDA.
App authentication, database, private storage, server functions, queues, and backups in the East US region.
Bounded AI processing for interpretation, drafting, explanations, planning, and safety classification.
App Store distribution, Sign in with Apple, Apple-controlled platform services, and purchases if subscriptions are offered.
Hosting and delivery of the public ARDA website.
Analytics, conversion measurement, and diagnostics on the public website only. ARDA does not send private app content to these website tools.
A production transactional-email provider must be selected and added to the public disclosures before email-code authentication is enabled for external users.