Artificial intelligence has helped Apple's human-based App Store Review process prevent over $2.2 billion in fraudulent transactions in 2025. It also rejected more harmful apps than ever before.

Policing the App Store for malware is a task for the App Store Review team, which is getting tougher over time thanks to AI-assisted app development. However, artificial intelligence is also helping Apple combat harmful versions of apps.

In a review of the 2025 submissions, Apple claims it had prevented over $2.2 billion in potentially fraudulent transactions. This brings the total over the last six years to more than $11.2 billion.

It also managed to reject over 2 million app submissions that could've been malicious and harmful to iPhone and Mac users.

The Trust and Safety Teams also stopped multiple large-scale attempts to create fraudulent accounts. Approximately 1.1 billion fraudulent customer account creations were detected and blocked, while 40.4 million customer accounts were deactivated over fraud and abuse.

Approximately 193,000 developer accounts were also deactivated in 2025 over fraud concerns. Another 138,000 developer enrollments were rejected.

Apple's security effort also extends beyond the App Store. It detected and blocked 28,000 illegitimate apps from pirate storefronts in 2025. This included not only pirated versions of legitimate App Store apps, but also gambling apps, adult content, and malware.

An AI assist

The continued growth of app submissions meant Apple had to develop its AI tools to cope. While human reviewers have worked with machine learning tools to handle the task, Apple has continued to invest more to improve those tools.

The AI elements are used to flag complex malicious patterns, analyze similar apps, and to determine potential issues in app updates.

This includes tactics such as apps that were approved but then modified post-review to facilitate financial fraud. Almost 59,000 apps were removed for bait-and-switch actions like that.

Over 22,000 submissions were also rejected for having hidden and undocumented features, 443,000 were rejected for privacy violations, and over 371,000 for copying other apps.

Another 2.5 million submissions were blocked from TestFlight for fraud and security issues.