Apple has rewritten several sections of its App Review Guidelines to more precisely target applications that use artificial intelligence to generate intimate images of individuals without their consent. The update takes aim at so-called “nudify” apps—software capable of altering real photographs to simulate nudity or sexualized content—while also expanding the tools Cupertino can use to crack down on apps that conceal problematic features until after receiving store approval.
The change does not introduce a ban out of thin air: App Store rules already allowed the rejection or removal of harmful content. However, Apple chose to explicitly name non-consensual intimate imagery among the risks that platforms hosting user-generated content must address. It is a more direct phrasing that leaves less ambiguity regarding what is expected of developers and what conduct could expose them to penalties.
For the apps affected, purely reactive moderation will no longer suffice. The guidelines state that anyone developing services where users can create, upload, or share material must implement adequate preventive measures and take action on identified violations. Failing to include sufficient safeguards, or failing to remedy issues following a report, can be deemed a violation of the Developer Code of Conduct.
This is a significant shift because it moves the focus from individual removed content to product design. If an app provides tools that make the creation of non-consensual sexualized deepfakes foreseeable, Apple can evaluate not only the after-the-fact response, but also the effectiveness of the safeguards put in place by the developer: filters, reporting procedures, audits, and restrictions on features enabling abuse.
From the Grok controversy to a more explicit rule
The revision comes after months of pressure on the company. Earlier this year, X users used Grok to generate non-consensual sexualized images of women and minors. The incident prompted lawmakers and advocacy groups to demand that Apple take action against X and the xAI chatbot, both available through the App Store.
Apple has not removed X or Grok from its digital storefront. Later, however, it confirmed to US senators that it had privately warned xAI that Grok could be banned from the App Store if the issue were not addressed. During June's WWDC, protesters also called on Apple to remove X, Grok, and nudification technologies from the store.
The debate does not solely involve major platforms. In January, the Tech Transparency Project identified dozens of AI apps linked to this trend on the App Store; a follow-up investigation in April reported that search suggestions and store ads could direct users toward similar applications. In July, San Francisco authorities ordered Apple to take down apps capable of generating non-consensual intimate imagery. Apple stated that it had removed the flagged applications.
The problem, therefore, is not merely the availability of an external generator accessible via the web. The App Store serves a function of distribution, discovery, and trust: an app published in the store can easily reach everyday users, including young people, and present itself with the veneer of a legitimate service. The new guidelines draw a sharper line between permitted creative tools and products that facilitate the abuse of someone's likeness.
The issue of features that appear after review
The other facet of the update concerns apps that behave differently from what was disclosed to Apple. Cupertino has introduced a specific section, 2.5.19, to prohibit hidden, dormant, or undocumented features. Apps must make their capabilities clear both to users and to the App Review team; attempting to conceal code or features through software obfuscation or other methods can result in removal from the App Store and the termination of the developer account.
The clarification targets the “bait-and-switch” model: an app is submitted for review with a seemingly harmless feature set and, once it clears the check, enables different or dangerous tools. Apple explicitly cites nudification among the examples of harmful capabilities that can surface after approval.
The ban on undisclosed features already existed in the policies, but the new wording specifies both concealment techniques and potential consequences. This is an important distinction for services that rely on remote servers for part of their functionality or update app behavior without shipping a new version: initial binary review alone is not always enough to catch subsequent changes.
The issue had also emerged in a different context. An investigation by 9to5Mac documented apps masquerading as games or utilities that, in Brazil, could display gambling platforms based on the user's IP address. The examples differ in content and intent, but point to the same limitation of traditional checks: an app can show one face during initial review and another under different geographic or operational conditions.
Penalties and the limits of the new policies
Apple has also strengthened its language regarding applications that facilitate human trafficking or child exploitation. These will be removed from the App Store and blocked from alternative distribution. The reference to alternative distribution carries weight in the European ecosystem, where rules introduced by the Digital Markets Act have opened up, under certain conditions, to alternative marketplaces and web distribution.
The updated rules therefore provide a broader basis for intervention, but do not automatically solve the problem. Determining whether an app has “appropriate” safeguards requires assessing how models actually work, how images are uploaded, what measures are in place against bypassing filters, and how quickly reports are handled. In addition, generative services can evolve rapidly through updates to models and cloud infrastructure.
The question of consistent enforcement also remains open. Guidelines can make it easier to reject a new app or challenge a developer, but their effectiveness will depend on the ability to identify products that change their behavior after entering the store and to act without waiting for abuse to become a public scandal. Precedents from recent months show why Apple chose to make its expectations explicit: the absence of clear phrasing leaves room for interpretation, while for victims, harm can occur and spread in minutes.
For developers working with image generation or manipulation tools, the update serves as an operational warning. It is not enough to state that unlawful use is prohibited in the terms of service: Apple demands effective safeguards and the remediation of reported violations. For users, the change does not amount to an absolute guarantee that no abusive app will ever reach the iPhone, but it raises the level of accountability required from services that host or generate content and strengthens the grounds for their removal.
