On October 16, 2026, Apple updated its App Store Review Guidelines to explicitly require developers to deal with nonconsensual sexual imagery, including images created by AI. Apps that digitally "undress" real people have already been banned, but the revision adds requirements to prevent and remove such images, and it again targets the tactic of enabling harmful features after an app has passed review. Penalties range from removing an app to banning the developer. However, the responsibility the rules impose differs between managing generated content and misleading the review process.
Three provisions, three different targets and penalties
The revision separates three decisions: what content users can generate or post, what happens when features change after review, and whether alternative distribution is allowed. Sorting the safety changes Apple announced against the relevant provisions of the Review Guidelines as revised on October 16 gives the following picture.
| Provision | Target and required developer response | Stated consequence |
|---|---|---|
| 1.2: User-generated content | Harmful content, including nonconsensual sexual images, AI-generated ones among them. Developers must have measures to prevent and remove it, and must fix violations that are flagged | The app may be removed depending on the response. Serious or repeated violations can lead to immediate removal and expulsion from the Developer Program |
| 2.5.19: Hidden features | Features not disclosed in review, or harmful features that appear after approval. Developers must make clear to reviewers and users what the app does | Removal of apps with hidden features and expulsion of the developer. Concealing features within the binary is also grounds for account termination |
| Section 5: Egregious exploitation | Apps that facilitate human trafficking or child exploitation | Notification of authorities, removal from the App Store, and blocking of alternative distribution |
The table organizes the three changes in Apple's announcement by target, required response, and consequence. It does not show records of individual app removals; it reflects how the rules treat violations depending on their nature and whether they are corrected.
Notably, the blocking of alternative distribution is explicitly stated only in Section 5, which covers human trafficking and child exploitation. That explanation cannot be read as a blanket measure applying to all undressing apps. Section 1.2, which requires action on AI-generated sexual images, and Section 5, which stops the circulation of apps that facilitate serious exploitation, need to be read separately according to what they cover.
Generative AI apps must also have filters and reporting
The revised Section 1.2 applies to apps that let users generate, post, or share content, and it explicitly names AI content-creation apps. Apple's announcement calls for "proactively preventing and removing" nonconsensual sexual images, including those generated by AI. The rule is not structured so that using generative AI releases an app from responsibility for managing its content.
The rule lists four measures. Apps need filters that prevent harmful content from being posted, and a mechanism to report offensive content and receive timely responses. They must also be able to block abusive users, and publish contact information that users can easily find. Providing a feature that creates images and running a contact point that responds after problems arise are treated as responsibilities of the same app.
There are also conditions for how developers respond when a violation is flagged. Under Section 1.2, Apple asks for the problematic content to be removed and for a plan to improve compliance. Depending on the developer's response, Apple may remove the app from the App Store until it shows improvement, and serious or repeated violations are grounds for immediate removal of the app and its developer.
What developers are asked to do here goes beyond writing prohibitions into their terms: they must have mechanisms that can actually prevent harmful content and handle reports. At the same time, the announcement does not specify implementation details, such as which detection technology to adopt or how to verify the consent of a person depicted in an image. The rule makes the required outcome clear, but whether technology exists that can achieve that outcome remains a separate question.
Swapping in features after review is also covered
Section 2.5.19 deals with the tactic of misrepresenting an app's purpose during review and revealing harmful functions, such as undressing, after approval. Apple describes this as swapping in features after review. It gives the example of adding functionality that was not in the version submitted for review, and says such apps are subject to removal from the App Store and the Apple Developer Program.
The ban on hidden features did not begin with this revision, however. Apple's notice of the revised items explains that the policy previously covered under 2.3.1 was clarified in the new 2.5.19. This is not a blanket prohibition on ordinary feature updates; what is at issue is hiding features from review or misrepresenting what an app does.
The text also covers attempts to conceal undisclosed functionality within the app binary, meaning the program that is distributed. Examples given include code obfuscation, which makes a program hard to read. Hiding features this way is grounds for removing the app and terminating the developer account.
Read together with 1.2, this provision shows that developers need both a system for managing what their apps generate and a system for accurately describing the functions they submitted. Setting up a channel to remove problematic images does nothing to address hiding harmful features during review. Conversely, an accurate submission does not end the responsibility for content management and report handling after distribution. Safety measures cannot be judged only at the moment an app passes review.
Read the published figures separately from the effect of the revision
In the same announcement, Apple said that in 2025 it prevented more than 150,000 apps with safety issues from reaching customers, and terminated more than 193,000 developer accounts involved in fraud or abuse. The first figure counts apps and the second counts accounts, so they cannot be added together as a total of apps removed. Both are figures for the company's overall safety efforts, not statistics counting only undressing apps.
These numbers do not allow a judgment about how much nonconsensual sexual imagery the revision will reduce. The announcement gives neither the number of undressing apps removed in connection with the revision nor the detection accuracy of the measures. Between writing the rules down and showing that harm has been prevented, a stage remains in which actual enforcement must be checked.
Developers remain responsible for dealing with harmful generated output and for managing the functions they present in review. In evaluating Apple's response, the useful evidence will be which apps are corrected or removed after violations are flagged, and what action is taken against feature swapping.