Privacy Policy

Prism is operated by Tradestreaks LLC ("we", "our"). This policy describes how we handle information when you use Prism on iPhone, iPad, Mac, Android, or the web app at app.prismpro.ai.

Information We Collect

• Apple Sign In and Google Sign-In: We receive the platform account identifier needed to authenticate you and, if you choose or the platform provides it, your name and email. Apple and Google accounts are not linked by email alone.
• Prompts and linked pages: Text you submit for AI comparison is sent to our servers and to selected AI providers (OpenAI, Anthropic, Google, Perplexity, and xAI) to generate responses. If your prompt contains a public web link, our server may retrieve that page so each selected model receives the same source material. Pages that require JavaScript may be processed by our isolated browser-rendering service.
• Attachments: Images, PDFs, and other files you attach are uploaded to our servers and forwarded to the models that can read them. See *Data Storage and Retention* for how long they are kept.
• Image generation: When you generate or compare images, your image prompt and any reference images you supply are sent to the selected image providers (OpenAI, Google, and xAI). Generated images are returned to your device and, for saved runs, stored with the run.
• Saved preferences and context: If you choose to save "About me" or response-style preferences, we store them with your account and send them to the models answering your requests. Separately, when history is enabled, our server extracts short summaries ("context documents") from your chats, Panels, and project files so later requests can carry relevant earlier material. Prism's synthesis/judge model does not receive saved preferences. Incognito chats neither use nor contribute to preferences or context documents.
• Location, only if you turn it on: Location sharing is off by default. If you enable it, Prism sends either a coarse place name (approximate mode) or device-rounded coordinates (precise mode, a separate opt-in) so answers and live research can be relevant to where you are. What was shared for a given answer is recorded in that answer's receipt. You can turn location sharing off at any time in Settings or in system settings.
• Camera, photo library, microphone, and speech: The camera is used only when you attach a photo. Photo-library access is scoped to the photo or file actions you choose. If you use voice input, audio is processed by the platform speech-recognition service to produce text; we receive only the resulting text, never the audio.
• Notifications, Live Activities, and Android completion alerts: If you allow notifications, we store the platform push token or Firebase Cloud Messaging registration for your device so we can update a Live Activity or notify you when a long-running comparison, Panel, Chat, or image run finishes. Push payloads carry run status and progress, not your prompt or the generated answer.
• Usage data: We track signed-in ask counts by plan and pool (Standard or Frontier) across comparisons, Panels, and chats, plus image-generation counts and the lifetime guest sample count, to enforce allowances.
• Guest access and retry data: If you use Prism without signing in, the app supplies platform integrity signals before guest asks are accepted. On Apple devices, this uses Apple DeviceCheck and App Attest. On Android, Prism uses Google Play Integrity when Android guest access is available. Our server verifies these signals with the platform, stores only the bounded device or integrity reference needed for abuse prevention, uses your IP address for rate limiting, and combines the verified reference with a client-generated request identifier to enforce the lifetime guest allowance and safely replay an ask after a lost network response. We do not use Apple's vendor identifier (IDFV) for guest enforcement.
• Purchase data: Subscription status is verified through Apple's In-App Purchase system or Google Play Billing, depending on where you subscribed. Apple and Google process payment details under their own terms; Prism stores only the product, transaction or purchase-token identifiers, entitlement state, and expiry information needed to activate and maintain your plan.
• Product analytics and diagnostics: We collect a limited vocabulary of feature interactions (for example, sign-in completion, paywall views, and result counts) plus Apple MetricKit crash, hang, CPU, and disk-write diagnostics. Analytics events use a random identifier that changes every app launch and are not linked to your Prism account. They never include prompts, responses, attachments, dictation text, location, email addresses, or persistent device identifiers.
• Content-safety reports: If you report a comparison, Chat response, or generated image, we store the selected safety category and the identifiers needed to locate the reported record. Optional report comments remain in Prism's feedback store. We send Better Stack a separate content-free operational alert containing only the category, report identifier, target type, and applicable run or candidate identifiers. That alert never includes your prompt, generated text or image, optional comments, email address, Apple identifier, or Prism account identifier.

How We Use Information

• Run AI model comparisons, Panels, chats, and image generations, and return synthesized results
• Authenticate your account and sync your history across your devices
• Enforce Guest, Free, and Pro plan allowances
• Deliver completion notifications, Live Activity updates, and Android run completion alerts you have opted into
• Improve reliability and support
• Review and respond to content-safety reports
• Understand aggregate feature use and diagnose crashes or performance failures

Permissions We Request

Prism requests only these, and only at the moment you use the feature: camera (attach a photo), photo or file access for the items you select, photo-library add (save a generated image), microphone and speech recognition (dictate a prompt), location (only if you turn on location sharing), and notifications (completion alerts and Live Activities where supported). We do not request access to contacts, calendars, health data, or broad photo-library reading.

Data Storage and Retention

• Account, comparison history, chat history, and Panel history are stored on our servers when you are signed in.
• Saved preferences remain with your account until you delete them in Settings or delete your account. Turning "Use saved preferences" off stops applying them but does not delete the saved text. Context documents extracted from your content are deleted when the content they came from is deleted, and with your account.
• Incognito chats are not added to local History, synced to your account, or eligible to become a saved comparison; saved preferences are not applied and no context documents are extracted. Their prompts and replies are still processed by our server and the selected AI provider. A completed reply may remain in the same account-, endpoint-, and request-scoped retry cache used by standard chats for up to 24 hours, solely to reconnect safely and avoid a duplicate charge.
• Guest comparisons and chats are not added to server history. When a guest ask includes a retry identifier, its typed response (including the comparison prompt/result or chat reply) remains available in an endpoint- and device-scoped replay cache for 24 hours so retrying does not spend another guest sample. Replay access expires at 24 hours; a retention job deletes expired cache records on startup and every six hours. Abandoned in-progress records expire after a shorter lease and their sample reservation is refunded.
• Signed-in comparison and chat asks use a client-generated retry identifier to prevent duplicate charges after a lost response. Account-, endpoint-, and request-scoped retry records and completed chat replies remain available for up to 24 hours. Abandoned in-progress records expire after a short lease and refund their exact reserved allowance. Completed comparisons also remain in normal History until you delete them, they reach the active-content retention limit, or you delete your account.
• The hashed guest device reference and used-sample count remain while needed to enforce the lifetime guest allowance. Rate-limit records expire with the configured rate-limit window.
• Comparisons remain in active server history for up to 365 days from creation. Chats and Panels remain for up to 365 days after their latest activity. You can delete them sooner at any time.
• When you delete a comparison, chat, or Panel — or it reaches the active-content limit — it leaves server history immediately. For chats and Panels, content-free synchronization tombstones may remain for up to 90 days so an offline device cannot restore deleted content.
• Feedback you provide on a comparison, Chat response, or generated image (ratings, content-safety categories, flags, optional comments, and in-app interaction signals such as copying an answer) is stored to improve result quality and review safety concerns. Feedback is kept until you delete the related content or your account and is purged with it. Content-free Better Stack operational alerts are kept only as long as reasonably needed for safety review and service reliability.
• Local device cache may store recent comparisons, chats, and Panels for offline viewing; deleting an item in Prism removes that local copy.
• Uploaded files begin as short-lived send inputs. Discarded staged files are deleted immediately when possible, including when an incognito chat is left or backgrounded; abandoned staged uploads expire after two days. When a file is used in a saved Chat, Panel, or comparison, its owner-scoped server copy follows the same bounded active-content retention window as that history so the attachment can be reopened across your devices. Project files remain until removed from or with their project.
• Generated images are retained with the run that produced them and are deleted when you delete that run or your account. Images you save to your photo library are yours and are outside our control.
• Pending analytics and diagnostics are held in a small, encrypted-at-rest on-device queue until delivery. Delivered records are kept in restricted operational logs only as long as reasonably needed to analyze product reliability and aggregate feature use.
• Retrieved public-page text may remain in a bounded server-memory cache for up to 24 hours to keep retries and multi-model comparisons consistent. It is not added to account history; URL paths and query strings are not written to operational logs.
• Deleting your account removes your conversations, profile and feedback from active systems; residual copies of that content are purged within 90 days. We retain purchase and usage accounting needed to honor subscriptions and prevent account deletion from resetting paid allowances. For an Apple purchase linked to a deleted registered account, we also retain a per-purchase keyed fingerprint of the verified sign-in identity for up to 400 days, or until billing access is recovered, whichever comes first. This pseudonymous billing record contains no email, raw sign-in identifier or conversation content. It lets older app versions restore billing after signing in with the same identity; it cannot restore deleted content. Expired fingerprints are excluded immediately and removed by the scheduled maintenance job within six hours.
• You can delete your Prism account in the app from Settings → Account & Legal → Delete account on Android, or from the account controls in Settings on iPhone, iPad, and Mac. If you no longer have the app installed or cannot sign in, email support@tradestreaks.com from the email address associated with your Prism account and ask us to delete your account. Account deletion does not cancel an Apple App Store or Google Play subscription; manage or cancel billing with the store where you subscribed.

Third-Party Services

Prompts, attachments, and image prompts are processed by third-party AI providers under their respective privacy policies. Public links that require JavaScript may also be processed by our browser-rendering infrastructure solely to retrieve the page content needed for your request. Prism routes some requests using a classifier that runs on a third-party model provider; the text sent to the classifier is the prompt you submitted. Apple and Google process platform sign-in, store billing, speech recognition, push delivery, and integrity checks where those platform services are used. We use Better Stack as an operational monitoring and incident-response provider. Prism's Better Stack source is configured to store the content-free safety alerts described above in Europe. Better Stack does not receive the reported prompt, generated output, optional comment, or Prism account identifier through this alert channel.

Contact

support@tradestreaks.com · prismpro.ai/privacy · prismpro.ai/account-deletion

Copyright © 2026 Tradestreaks LLC. All rights reserved.