Riya AI
Security

Built to be trusted with your customers' conversations

Plain answers to the questions businesses ask before letting an AI talk to their customers.

Where data is stored

Workspace data, contacts, transcripts and summaries are stored in Google Cloud Firestore in Mumbai (asia-south1). The voice engine runs on a server in Mumbai and handles call audio live; the audio is not stored. Some processing happens outside India: the language model (Anthropic, USA), and Vercel (a US company), whose functions run in Mumbai but whose pages are delivered through a global network. The privacy policy lists each provider.

Encrypted in transit and at rest

Every connection uses TLS 1.2 or later. Storage is encrypted at rest by Google Cloud. Secrets such as API keys and tokens are stored hashed where possible and never shown again after creation.

Workspace isolation

Every record carries its workspace id and every query is scoped to it. Owners, admins and members have distinct permissions. Workspace API keys can be created and revoked at any time and are rate limited.

Personal identifiers masked

Card numbers, Aadhaar and PAN-like strings and OTPs are masked in transcripts before storage. Outgoing webhooks exclude transcripts unless you turn them on per webhook.

Disclosure on every call

The agent introduces itself as an AI assistant of your business, and the standard openings tell callers that a written record of the call is kept. No call audio is recorded. Outbound calls run only between 9 AM and 9 PM IST, per TRAI rules. A do-not-call list is honoured automatically.

Retention and deletion

Transcripts are deleted after 90 days by default; we can set a shorter period for your workspace on request (7 days at least). No call audio is stored. Owners can delete any workspace and all its data from Settings at any time. Deletion requests from your customers are honoured on your instruction, as required by the DPDP Act 2023.

Processor, not controller

Riya AI processes call data only on your instructions. You remain the data fiduciary for your customers. A separate data processing agreement can be agreed with you in writing.

Sub-processors

Plivo (telephony), Oracle Cloud in Mumbai (the voice engine server), Sarvam AI, India (speech to text and voice), Anthropic, USA (language model; no training on your data under its API terms), Google Cloud Firestore in Mumbai (database), Vercel (website and dashboard hosting), Meta (WhatsApp Business Platform, only where switched on). The privacy policy says what each one receives.

Reporting a problem

Found a security issue? Send it through the contact form on this website with "security" at the start, or call or WhatsApp +91 94668 55953. We acknowledge within one working day and do not take action against good-faith research.

Details on what we collect and why are in the privacy policy. Contractual terms are in the terms of service. Last updated 17 November 2026.

Put Riya on the phone this week

Free workspace, test by typing, pay only for connected minutes once you go live. Design partners in Haryana and Delhi NCR get hands-on setup from our team.

Questions? WhatsApp or call +91 94668 55953, or use the contact form.