Can I trust ChatGPT?

Trust checks

• Secure by Design

  ● Strong

  Strong security foundation

  Why this result

  This check looks for security issues at high severity or above in the code of this build, and none were recorded. Its scope covers stored credentials, how the app protects data on the device, and how it protects data in transit.

• Data Minimization

  ● Strong

  Tracking done right

  Why this result

  The check looks at analytics and attribution code for companies other than the developer: the data it reads, the companies named in code as recipients, and whether a consent step runs before that code. In this build, either no such code was recorded, or a consent step runs before it and the app's own disclosures describe the collection.

  The app does not request the advertising identifier and has no cross-app tracking framework present. Install attribution uses Apple's privacy-preserving network-level system, which never exposes individual user identities to advertisers. Four analytics and crash-reporting services are linked for session and performance monitoring, but none involve advertising identifiers, and no evidence of data collection before a user engages with the app was found. The app correctly avoids the device-level tracking consent prompt because it does not access cross-app identifiers.

• Manifest Mismatch

  ◐ One criterion not met

  Disclosure incomplete or contradicted

  Why this result

  The check compares the app's privacy disclosures against what the code does, and at least one claim did not match in this build. Mismatches of this kind include a data category collected but not disclosed, a recipient the disclosure omits, and an identifier attached to data the disclosure describes as anonymous.

• User Control

  ● Strong

  No lock-in

  Why this result

  The check looks for a way to export the data an account has accumulated and a way to delete the account itself. In this build, either both were found, or the app has no account of its own and keeps its data on the device. For an app without an account, the developer has no server-side copy to export or delete, and deleting the app removes the data in its own storage.

  Conversation history and persistent memory are stored on OpenAI's servers, creating a server-side data store that the user cannot access independently of the platform. No in-app account deletion or conversation export flow is confirmed in the app bundle. On the other hand, no proprietary data formats or artificial switching barriers were identified, and the app uses standard web authentication rather than a custom identity system that would complicate leaving the service. Whether users can retrieve or delete their stored data through the app requires review of the live in-app settings.

What the app contains

• 21 services

  Third-party services the scan names in the build:

  • Auth0
  • Segment
  • Datadog dd-sdk-ios
  • Sentry
  • PLCrashReporter
  • RevenueCat (CustomEntitlementComputation)
  • Stripe (StripeCore, StripePaymentSheet, StripePayments, StripePaymentsUI, Strip…
  • Plaid LinkKit
  • Statsig
  • LiveKitWebRTC
  • Lottie
  • Kingfisher
  • PhoneNumberKit
  • swift-nio (NIOPosix)
  • swift-protobuf
  • Highlightr
  • KaTeX
  • SwiftCSV
  • OpenAPIKit
  • EmojiKit
  • OAIMarkdown

Changes since the previous scan

Method and limits

Static analysis: the decompiled code, manifest and resources of this build, read file by file. Network traffic at run time is established by a capture of the running app.

Scan date

11 August 2026

Corrections

No correction is published for this app.