Know where
your information goes.
This page describes the public INRI GPT website and demo. Business projects have their own agreed scope and data handling requirements.
Updated 3 October 2026. INRI GPT is created by Inri Inc, founded by Inri Bollo and Unejs Osmani.
The public AI demo
When you run an analysis, your selected workflow and text travel over HTTPS to our server and then to Cloudflare Workers AI. The service generates a response and returns it to your browser. INRI GPT does not save demo prompts or generated outputs in its application storage and does not use them to train models.
Cloudflare processes the request as a service provider. Its Workers AI data policy states that customer content is not used for model training without explicit consent. Use sample or redacted text; do not submit confidential records, passwords, financial account details, or patient information to this public demo.
Chat workspace, documents, images, and audio
Chat messages, recent conversation context, document excerpts, image prompts, attached images, and audio recordings are sent over HTTPS to our server and Cloudflare Workers AI when you press Send. Models include Qwen3 30B-A3B for text, FLUX.1 [schnell] for image generation, LLaVA for image understanding, and Whisper for transcription. The INRI GPT application does not persist these inputs or outputs on the server.
Chat text is stored in this browser tab?s session storage so you can switch conversations and refresh the page. Use Delete conversation to remove a chat. Attachments, generated images, and charts remain in page memory and are not restored after refreshing. Downloads you choose to save remain on your device. The color theme preference is stored locally in your browser.
Document text is extracted in your browser, with clear size and excerpt limits. PDF extraction reads up to 20 pages; only a bounded excerpt is sent for AI processing. Images are resized locally. Audio is converted locally to mono WAV with a 30-second limit. Your microphone is accessed only when you choose Record and grant permission. Browser read-aloud uses the speech service and voices available on your device, which may depend on your browser or operating system provider. The calculator and CSV chart calculations run locally without model requests.
Recent chat context is limited to five exchanges and the current message, within a 12,000-byte limit. Source attachments add up to 16,000 bytes. All AI tools require Turnstile verification and are subject to usage caps. INRI GPT does not share your chat content with Unays LLC through the partnership links.
Customer accounts, organizations, and purchases
We store your account name, email, password hash, organization memberships, purchase records, project briefs, messages, uploaded files, shared company documents, saved organization conversations, workflow templates, and AI usage metadata on our Oracle-hosted server. Project information and organization workspace content are visible to authorized members and Inri Inc administrators. Saved organization conversations are distinct from the public chat: conversation messages and source excerpts remain on our server until deleted. Knowledge answers send selected matching document excerpts to Cloudflare Workers AI. AI usage records contain request metadata, not a copy of prompts or outputs. Payment details are processed by PayPal, or by Paddle when its card option is available; we store order references, amounts, currencies, and payment status, not card details. Passwords are hashed with Argon2id. API keys, invitation tokens, and reset tokens are stored as hashes.
Google Workspace linking uses only identity permissions. We store the linked Google account identifier, verified email, and Workspace domain without keeping Google access or refresh tokens. Customer session cookies are Secure, HttpOnly, and SameSite=Lax. Contact us to request account or project-data deletion; purchase records may need to be retained for accounting obligations.
Advertising
Selected workflow articles may request Google AdSense advertising. Chat, interactive guides, demos, navigation, legal, contact, account, checkout, dashboard, and administration pages do not load AdSense advertising. Google Privacy and messaging handles any required regional consent. Private account, checkout, dashboard, and administration pages do not load advertising. Google may process technical information and use cookies or local storage when ads are served; see its advertising policy. Our advertising code does not submit AI prompts or private account fields to Google. Inri Inc must configure Google Privacy and messaging where a certified consent platform is required.
Project inquiries
The contact form collects the name, email address, optional company name, and message you choose to provide. We use these details to review and respond to your inquiry. Submissions are stored in a restricted folder on our Oracle-hosted server, outside the public website directory, and are automatically removed from active inquiry storage after 180 days. Local backups retain up to three backup versions.
Abuse prevention and technical logs
Cloudflare Turnstile checks that form requests come from legitimate visitors. Cloudflare receives technical information needed for this verification; see its privacy policy. Our server uses IP-derived identifiers for short-term request limiting. Normal hosting and security logs can contain IP addresses, request paths, and timestamps. Request bodies are not deliberately logged by the application.
Google advertising can use technical storage as described above. We do not add separate analytics cookies. Turnstile may use technical browser storage or signals necessary to perform its checks. The workflow choice can be held temporarily in your browser session when following a demo link.
Security and production requirements
We use HTTPS, server-side secrets, limited input sizes, Turnstile verification, request throttling, and a daily demo quota. The model cannot run code, browse your systems, or take actions for your business. Outputs can still be incorrect and require human review.
We do not claim SOC 2 certification or blanket regulatory compliance. Data residency, access controls, contractual terms, retention, and any regulated-data requirements must be reviewed for a production project.
Questions or deletion requests
Use the contact form to request access, correction, or deletion of an inquiry, or to ask about data handling. Include the email address used for the original inquiry so we can verify the request. Do not include sensitive identity documents in the form.
Optional Google Drive synchronization
Organization owners can grant read-only Drive access and select files for synchronization. Only selected Google Docs and supported text files are copied into their shared library. Offline connection tokens are encrypted on the server. Source text may be sent to Cloudflare Workers AI as relevant excerpts when a user requests an answer. Disconnecting removes the stored Drive connection and synchronized library copies; existing saved conversation excerpts and retained recovery backups remain until deleted or expired. We do not edit original Drive files.
First-party aggregate measurement
We count demo submissions, pricing views, created checkouts and verified purchases by day to understand the purchase funnel. These aggregate records contain no visitor identifier, email, prompt, document, IP address, query string or browsing profile and expire after 90 days. Browser Do Not Track and Global Privacy Control preferences suppress browser-side events. Necessary payment records and ordinary security/server logs are handled separately.
Recovery copies
Encrypted recovery snapshots are retained for seven server copies and fourteen off-server copies on the owner's computer. Selected document text and account records may be included. Backups are used for recovery and tested in an isolated database; recovery keys are stored off the production server.
Projects, source history and website support
Organization members can save project briefs, assigned tasks, comments, work outputs and review decisions. These are shared with authorized organization members. The library retains up to twenty distinct document versions; deleting a document removes its version history, while excerpts already saved in conversations and project outputs remain until those items or account data are removed.
When an organization enables human support in its website widget, a visitor may explicitly submit their name, email and message. Conversation messages are visible to the organization team. A random access token, stored in that browser tab, allows the visitor to read and reply to their own conversation for seven days. Tokens are stored as hashes on the server. The organization owner can delete a conversation; no automated email is sent. Widget messages are not added to company knowledge or sent to AI by the support feature.
Public business-kit briefs and downloads are generated locally in your browser and are not sent to our server. Support conversations and projects may be included in encrypted recovery copies described above.