Apple's App Review Guidelines now name AI nudify apps, and new rule 2.5.19 bans hidden featuresApple's App Review Guidelines now name AI nudify apps, and new rule 2.5.19 bans hidden features

Apple revised its App Review Guidelines on 16 October, adding nonconsensual intimate imagery to the rule on user-generated content and creating a new rule against hidden app features.

The changes were published in two Apple Developer News posts five minutes apart. The first explains the reasoning and names "nudification" apps, which Apple describes as apps that create sexually explicit images of real people without their consent. The second is the formal changelog, which lists seven edits to the guidelines. Translations follow within a month, Apple says.

What changed in rule 1.2?

Guideline 1.2 covers apps where people create, generate, upload or share content. Its opening sentence used to list intellectual property infringement and anonymous bullying as the risks; it now lists nonconsensual intimate imagery between them. The rule also now spells out which apps it covers: social media, user-generated content platforms and AI content creation apps.

The four safeguards such apps must have are unchanged: a filter for objectionable material, a way to report content with timely responses, the ability to block abusive users, and published contact details. What is new is the consequence. Developers who do not put appropriate safeguards in place, or do not fix violations after Apple tells them, "may be considered in violation of the Developer Code of Conduct", the guideline now reads. Apple's post says that risks removal from the App Store.

Apple's post also restates that apps generating nudity or sexually explicit imagery were already banned, and that it removes them when it finds them.

What does the new rule 2.5.19 cover?

Guideline 2.5.19 says apps must not include hidden, dormant or undocumented features, and that an app's functionality should be clear to users and to App Review. Hiding features in the binary, including through code obfuscation, is grounds for removal from the App Store and termination of the developer account. The changelog says the policy is long-standing and was previously part of guideline 2.3.1, which deals with accurate metadata.

Apple's post ties it to what it calls bait-and-switch: apps that misrepresent their purpose during review and reveal capabilities such as nudification only after approval. Neither post names an app or a developer.

The smaller changes

• Section 5: apps found to facilitate human trafficking or the exploitation of children will be removed from the App Store and now also blocked from alternative distribution outside the App Store.
• 2.3.14, new: creative assets, the header images and video developers can now use on product pages and in search, must relate to the app or its brand, suit all audiences, and carry no specific or discounted prices.
• 3.1.1(a): the separate bullet on the Music Streaming Services Entitlement is gone, folded into the bullet on StoreKit External Purchase Link Entitlements, which now says the link goes to a permitted destination.
• Before You Submit and After You Submit: links to further Apple documentation, which is the change the shorter coverage of the revision led with.

Rule 2.3.14 is the review side of the header images and looping video Apple opened to App Store product pages at the start of the month.

Why now?

Apple's posts do not say, but the year gives context. In January Apple privately warned xAI that Grok could be removed over sexualised deepfakes, according to its 30 January letter to three US senators, which NBC News reported in April; a later version of Grok was approved. In July, San Francisco's City Attorney sent Apple and Google cease and desist letters over nudify apps, and Apple said it had removed three of the apps named, MacDailyNews reported. 9to5Mac also points to its own July finding of gambling apps that showed their betting features only to users in Brazil, which is the pattern 2.5.19 describes.

Apple's post gives two numbers for last year: more than 150,000 apps with safety issues stopped before release, and more than 193,000 developer accounts terminated for fraud and abuse.

Our take

Read closely, rule 1.2 changes less for apps than for developers. The four safeguards are the same four the rule has asked for since the early App Store. What has changed is that failing to act after a warning is now framed as a Code of Conduct matter, which is about the developer account, not one binary. That moves the risk from one rejected update to everything a developer publishes.

Naming AI content creation apps alongside social networks matters more than it looks. An app that generates images was never obviously a user-generated content platform in the old sense; it now plainly is, so it needs the report button, the block option and the contact page. Those are interface obligations, and for a small image or avatar app they are the parts most often left until last. From a product point of view, a report flow that reaches a human in reasonable time is the safeguard that does the most work here, and it is the one an app can be judged on after notice.

Rule 2.5.19 is a clarification, and Apple says so. Its value is that the rule now has its own number. The guideline numbers are what Apple cites when it removes an app, as it did when it told Congress that rule 1.1.1 was the basis for removing ICEBlock. A bait-and-switch removal can now point at a rule written for exactly that.

What is still unknown is how App Review will find features that are built to stay hidden during review. A guideline does not detect a server switch that turns a feature on later; that takes reports from users and outside researchers, which is how most of this year's cases came to light. The trade-off APPDOOK sees is that the rules are now explicit while detection still largely depends on someone outside Apple noticing first.

What should app makers check?

Any app that lets people generate or share images should check that its reporting, blocking and contact paths actually work and get answered. Apps that switch features on remotely should make sure those features are described in the review notes, as 2.3.1 already required. And anyone planning header images or video should read 2.3.14 before submitting them: discount prices in the artwork are now ruled out in the guidelines themselves.