The App Store Review Guidelines were updated for AI content on October 16, 2026. In a post on its developer news feed that day, Apple explained how the revision deals with nonconsensual intimate imagery and with apps that swap in new capabilities after passing review. This is not a brand-new prohibition: Apple stresses that pornographic material, "nudification" apps that turn photos of real people into nude or sexualized images with AI, and hidden features were already banned. Three things actually change — AI content creation apps are now named as subject to content moderation duties, nonconsensual intimate imagery must be proactively prevented and removed, and the cost of violations is spelled out at the level of the Developer Code of Conduct and account termination.
What Guideline 1.2 changes: generators now stand on the same line as social platforms
Guideline 1.2 covers user-generated content. After the revision, the content moderation requirements explicitly cover nonconsensual intimate imagery, including imagery generated using AI tools, as content that apps must proactively prevent and remove; the guideline text also names AI content creation apps alongside social media and user-generated content platforms. The existing duties — filtering objectionable material, offering a way to report content, and letting users block abusive accounts — remain in force. The new consequence is that developers who fail to implement adequate safeguards, or who do not remedy violations after notice, risk violating the Developer Code of Conduct and being removed from the App Store. The risk is not abstract: earlier this year, xAI's Grok was widely used to generate sexualized deepfakes of other people, and the UK Information Commissioner's Office then made xAI the subject of a separate formal investigation (see the UK ICO pressing major model developers on data practices). Apple has now written "generators are responsible for what they produce" into its general rules instead of handling it case by case.
New Guideline 2.5.19: passing review is not the finish line
The new Guideline 2.5.19 clarifies a longstanding prohibition: an app must not misrepresent what it does during review and then change after approval, for example by adding features that were not in the submitted version. Apps caught doing this are removed from the App Store, and their developers are removed from the Apple Developer Program. The guideline text also makes clear that hidden, dormant, or undocumented features are not allowed in themselves, and that concealing them through code obfuscation counts as a violation too. In development terms, quietly switching on an undeclared generation capability after approval through a server-side flag or remote configuration falls squarely within this rule; passing review is now only the start of compliance, not an exemption period.
Section 5 and the scope boundary: a global platform rule, not a national law
The Section 5 change targets the gravest category: apps that facilitate human trafficking or the exploitation of children will not only be removed from the App Store but also blocked from alternative distribution — in regions that allow third-party app marketplaces, this closes the escape route of simply relisting elsewhere. The boundary matters: these guidelines are Apple's platform rules, not any country's law. They bind developers who distribute through the App Store worldwide as a condition of distribution, and they do not directly govern the same app on the web or on Android; but in most markets iOS distribution cannot bypass the App Store, so their practical force is considerable. Apple also disclosed that over the past year it prevented more than 150,000 apps with safety issues from reaching customers and terminated more than 193,000 developer accounts engaged in fraud and abuse.
Four fixes for anyone shipping a generative app on iOS
First, inventory the entry points: any feature that accepts photos of real people or can generate likenesses of real people should be treated as a high-risk surface, not something to inspect only after complaints arrive. Second, move the defense upstream: sexualized editing of real people's images has to be blocked before generation, because takedown alone no longer meets the literal "proactively prevent" requirement, and reporting and blocking tools must stay available. Third, build a response channel for Apple's notices: someone inside the team must own receiving notices, remediating within a set time, and keeping records, because failing to remedy after notice is itself now written in as a violation. Fourth, align feature flags with the reviewed version: remote configuration and gradual rollouts must not enable undeclared capabilities, and functional changes should go through the normal update process. For iOS-only teams, losing Developer Program membership is far more serious than a single rejection — it nearly shuts down the entire distribution channel. Now that the rules are written down, remediation and appeals at least have a clear reference point, but whether enforcement treats large platforms and small developers alike will only be visible in the removal cases that follow.
