Apple has revised its App Store rules to put a sharper, more explicit focus on AI-generated nonconsensual intimate imagery, commonly described as sexual deepfakes. The change matters because it does not merely restate the store’s existing ban on explicit material. It places an affirmative responsibility on developers of relevant apps, including AI content-creation apps, to prevent this material and remove it when it appears.
Apps that fail to do so risk being removed from the App Store. Apple also says developers can face removal from the Apple Developer Program when they violate the guidelines. In practical terms, the policy gives Apple a clearer route to act against both an offending app and, potentially, the developer account behind it.
The revisions arrive alongside a new rule aimed at another alleged route around platform review: apps that present one set of features to reviewers and add harmful functionality after approval. Apple specifically connects the clarification to so-called nudification apps, which generate sexually explicit imagery of real people without their consent.
A specific obligation for AI content apps
Deepfakes are synthetic or altered media made to depict a person saying or doing something they did not actually say or do. In this case, the policy focus is AI-generated intimate imagery made without the subject’s consent. “Nudification” is a more specific term for tools that transform an image of a real person into a fabricated nude or sexually explicit image.
Apple already prohibited sexually explicit content, but the revised guideline makes the AI angle unmistakable. Developers cannot treat harmful user-made deepfakes as an issue outside their responsibility simply because an app’s tools generated the material at a user’s request. The updated Rule 1.2 expressly names AI content-creation apps and says developers must proactively prevent and remove prohibited AI-generated imagery.
That word, proactively, is significant. A reactive system waits for someone to report material after it has been created or distributed. A proactive approach implies developers need safeguards intended to stop abuse before it reaches users, alongside a process to remove violations that do get through. Apple’s published update does not prescribe one universal technical solution, so it does not establish a single required moderation tool or detection method. It does establish the result for which developers are accountable: prohibited content must be prevented and removed.
The rule is broad enough to matter beyond a single kind of AI product. Image generators are the obvious example, but the wording’s reference to AI content-creation apps means developers should consider how their features can be combined, prompted, or misused to produce nonconsensual intimate material. The policy concern is the harmful output and an app’s failure to address it, rather than AI as a category.
Grok provides an important piece of context
The updated language could apply to apps such as Grok, xAI’s chatbot. Users had complained that Grok was being used to create nude and sexualized deepfakes. Earlier in the year, Apple told xAI that Grok would be removed from the App Store if it continued to allow nude deepfakes. xAI subsequently added safeguards, and Apple approved an updated version of the app.
That sequence helps explain why Apple has opted for more direct wording in the guidelines. A general prohibition on explicit content addresses what is visible in an app, but AI systems can create outputs dynamically from user inputs. Naming AI creation products and nonconsensual synthetic imagery narrows the room for uncertainty over whether these harms fall under a platform’s existing rules.
It is important not to overread the change. The revision does not identify one app as its sole target, nor does it say every AI tool creates this kind of content. Instead, it makes clear that an app offering AI-driven creation cannot ignore foreseeable misuse involving intimate depictions of real individuals. The Grok episode is relevant context for the tougher language, not proof that the new rule is limited to that product.
The post-review feature problem
Apple also added Guideline 2.5.19, which addresses functionality that was not part of the version submitted for App Store review. Under the clarification, a developer that adds features after review which were absent from the submitted app can risk removal from both the App Store and the Apple Developer Program.
The concern is particularly acute when an app conceals its actual purpose during review. An app might appear to be a conventional photo utility, for example, then reveal a harmful capability only after it has been approved. Apple characterizes this conduct as a bait-and-switch: presenting a benign version for review, then changing the proposition once the app is live.
Apps containing hidden or undocumented features are prohibited, including apps that misrepresent their purpose during review and reveal harmful capabilities only after approval.
This distinction is useful for understanding the policy. The deepfake provision deals with what developers must do about prohibited AI-generated intimate content. Guideline 2.5.19 addresses a separate enforcement challenge: whether Apple was shown the real app at all. An app may violate one rule by allowing harmful outputs, another by hiding the feature that creates them, or both.
For legitimate developers, the message is not that apps can never evolve after launch. Software routinely changes through updates and server-side functionality. The issue described by Apple is undisclosed functionality and misrepresentation during review, especially when the concealed capability is harmful. Developers adding material functions after approval have stronger reason to ensure those changes are properly disclosed and consistent with the app that was reviewed.
What this means for users and developers
For people using AI-enabled apps, the policy creates a clearer basis for expecting platforms to respond to sexual deepfake abuse. It does not mean harmful content can never be produced or that every moderation decision will be immediate. Rather, it states that developers are expected to maintain prevention and removal measures, while Apple reserves the ability to remove apps when those measures are absent or inadequate.
For developers, the change raises the stakes around product design, moderation planning, and release practices. Teams working on image generation, chatbots, creative editing, or other systems capable of producing user-directed media need to account for the risk of nonconsensual intimate outputs. Teams also need to avoid treating review as a hurdle that can be bypassed through undocumented functions or later feature activation.
- Prevention: apps covered by the rule need safeguards intended to stop prohibited AI-generated intimate imagery.
- Removal: developers must act to remove offending material rather than leave it available after becoming aware of it.
- Transparency: functionality presented to App Store review should accurately represent what the app does.
- Consequences: enforcement can extend beyond taking down an individual app to termination from the Apple Developer Program.
These are platform rules, not a complete description of every legal question that can arise from synthetic imagery. Apple’s policy is about App Store eligibility and developer participation. Its usefulness lies in making the app-distribution consequences more explicit, particularly where an app’s capabilities are hidden or where a developer does not meaningfully address prohibited use.
Alternative distribution is included, too
Apple’s update also changes Section 5 to extend its prohibition on apps facilitating human trafficking or child exploitation to alternative app distribution. That is a separate but related moderation development: the rule is not confined to the traditional App Store channel.
The expansion signals that Apple is applying certain high-severity safety restrictions across the ways apps may be distributed, rather than treating an alternative channel as outside those standards. The supplied guidance does not add operational detail about how that enforcement will work in every case, but the policy direction is clear: distribution method does not exempt an app from this particular ban.
Apple’s enforcement posture is direct. The company states that it will remove apps and terminate developers that violate its guidelines. Combined with the specific language on AI deepfakes, nudification, undisclosed features, and alternative distribution, the update gives developers less basis to claim that harmful post-review functionality sits in a policy grey area.
Why the wording matters
App-store policies often work as both restrictions and enforcement tools. Broad standards can cover many types of misconduct, but detailed examples make it easier for developers to understand what conduct is prohibited and for platform operators to explain why action was taken. Here, Apple has done both: it has identified AI-generated nonconsensual intimate imagery as a moderation responsibility, and it has identified the hidden-feature tactic that could be used to obscure nudification tools until after review.
The result is a rule set that concentrates on two linked problems. First, an AI app must not enable and leave up nonconsensual intimate deepfakes. Second, an app cannot seek approval under one description while reserving a harmful, undocumented feature for later. That combination is likely to be the practical takeaway for developers building AI tools for Apple platforms.
As platform rules increasingly shape the software business, those enforcement boundaries are part of the wider industry picture. For another example of the pressures affecting game and technology companies, see our coverage of Liquid Swords entering bankruptcy months after Samson launched.
