Apple has updated its App Store guidelines to ban apps that use artificial intelligence to create offensive content, especially those that generate images of people without clothes. The changes took effect on the weekend before October 19.
The most specific point concerns bait-and-switch: an app is approved as a harmless tool and only later introduces or unlocks harmful features, including AI-based nudification. Apple said it will remove apps that add features missing from the version submitted for review. The same rules require apps hosting user-generated content to filter unacceptable material, offer reporting tools and allow blocking abusive accounts.
Image: 2degrees.nz — https://www.2degrees.nz/shop/apple
The crackdown follows months of pressure on the phenomenon. In the past year, review blocked more than 150,000 apps from reaching users and led to the removal of 193,000 developer accounts for fraud and abuse. In July, Apple removed Kromix after WIRED found pornographic deepfake ads of a US politician; there too the app reportedly added banned features after its initial review.
The problem does not stop at apps, however. An AI Forensics report showed the underlying models are hosted unfiltered on Hugging Face, and removing an interface does not remove the underlying capability. The European ban on nudify apps takes effect on December 2, while an Agora Digitale Transformation analysis found 50 members of the European Parliament associated with sexual deepfake material across 200 sites.
Image: logo.com — https://logo.com/blog/apple-logo
The open question is how far review rules can stop hidden features. Apple says it can detect bait-and-switch, but the gap between what gets approved and what an app does later remains the hardest part to police.
No comments yet. Be the first!