Technology Providers & AI Disclosure

R3 is operated by one person and runs on managed infrastructure. That means some processing of client information happens through third-party providers. This page names them and describes what each one actually does in the current implementation.

Rishav Aryal (ABN 34 930 918 759) is the sole-trader service provider. R3 Racing Analytics is the service/brand name used on this site.

Supabase provides the authentication system (including multi-factor authentication for administrators), the managed Postgres database that holds client records, and the private storage buckets that hold uploaded documents and published reports.

Access to client records is enforced at the database level by row-level security policies, so a signed-in client can only reach their own client scope.

Lovable provides the application and deployment infrastructure that serves the R3 site and portal, and the managed transactional email service that sends authentication and notification email.

Some features can route AI requests through the Lovable AI Gateway rather than calling a model provider directly.

The R3 Assistant's model provider is configured on the server. The current code defaults to Groq, and can use the Lovable AI Gateway as an alternative or fallback path.

Because the provider is environment-configurable, R3 does not claim that one specific provider is always used. If the configured provider changes in a way that is material to clients, that change will be reflected in the Privacy Policy and customer disclosures.

The Assistant is assistive and read-only in nature. Fair use limits are described in the AI & Fair Use Policy.

Quick Add Race can optionally transcribe an uploaded race image or PDF. When that feature is explicitly enabled, the request is routed through the Lovable AI Gateway to Google Gemini 2.5 Flash.

According to R3's internal Quick Add AI privacy documentation, this transcription path is disabled by default. When it is disabled, no source material is sent to a model provider for transcription.

The R3 Assistant code does not intentionally store prompt or answer text. It stores operational metadata — quota counters, request status, latency and token counts — so R3 can run the feature reliably and enforce fair use.

Quick Add source material and transcription logging behave as described in R3's internal Quick Add AI privacy documentation and in the Data Retention Policy: rejected and abandoned drafts, and their source material, are designed to be purged after 30 days.

R3 currently uses AI only as an assistive tool. R3 does not intentionally use AI to make decisions that could reasonably be expected to significantly affect an individual's legal rights or interests. If that changes, this policy and R3's customer disclosures will be updated as required.

R3 cannot independently verify a model provider's internal retention or training practices, and does not make guarantees on their behalf.

What R3 can control and does control is the scope of what is sent: AI features are used narrowly, transcription is off by default, and prompt/answer text is not intentionally retained by R3.

Some of these providers process or store information outside Australia. Handling of personal information, including overseas disclosure, is described in the Privacy Policy.

Providers can change for legitimate operational, reliability, cost or security reasons. Where a change is materially adverse to clients, it will be reflected in the Privacy Policy and in customer disclosures.