Google disclosed this week that its artificial intelligence systems played a central role in preventing 2.36 million policy-violating Android apps from reaching the Play Store in 2025, a figure that underscores both the scale of the threat facing mobile users and the increasingly automated nature of the company’s defenses. The announcement, detailed in Google’s annual security report, represents a significant escalation in the cat-and-mouse game between platform operators and malicious developers who seek to exploit the world’s most widely used mobile operating system.
According to TechCrunch, Google said that AI-assisted reviews were responsible for more than 92% of the human reviews conducted on apps flagged for policy violations. The company noted that this AI-driven approach allowed it to act more quickly and at greater scale than would have been possible with human reviewers alone, catching threats that might otherwise have slipped through during periods of high submission volume.
The Numbers Behind Google’s 2025 Crackdown
The 2.36 million blocked apps represent a notable increase over previous years. In 2024, Google reported blocking approximately 2.28 million apps, and in 2023, the figure stood at around 2.28 million as well. The upward trend suggests not that Google’s defenses are weakening, but rather that the volume of attempted abuse continues to climb. Malicious actors are submitting apps at an accelerating pace, testing the boundaries of automated detection systems with increasingly sophisticated techniques including code obfuscation, delayed payload delivery, and social engineering tactics designed to fool both algorithms and human reviewers.
Google also reported that it banned 158,000 developer accounts in 2025 for attempting to publish malware or engaging in repeated policy violations. This figure is part of a broader enforcement strategy that goes beyond simply removing individual apps. By targeting the accounts behind malicious submissions, Google aims to raise the cost of doing business for bad actors, making it harder for them to simply create new accounts and resubmit harmful software under different names.
AI as the First Line of Defense
The role of AI in Google’s security apparatus has grown substantially over the past several years, but 2025 appears to mark a turning point in terms of the technology’s centrality to the review process. Google indicated that its machine learning models are now trained on vast datasets of previously identified malware, spyware, and policy-violating apps, enabling them to recognize patterns and flag suspicious submissions before they ever reach a human reviewer. The AI systems analyze everything from an app’s code structure and permission requests to its metadata, developer history, and behavioral signals observed during automated testing.
This approach has allowed Google to dramatically reduce the time between an app’s submission and its removal when it is found to be in violation. In many cases, flagged apps are now caught and blocked before they are ever published to the store, meaning users never encounter them. However, Google acknowledged that some malicious apps do still make it through initial screening, particularly those that employ techniques such as dynamic code loading — where the harmful functionality is not present in the app at the time of review but is downloaded later from a remote server after the app has been installed on a user’s device.
The Persistent Challenge of Sideloading and Third-Party Stores
While the Play Store remains the primary distribution channel for Android apps, a significant portion of the malware threat comes from outside Google’s walled garden. Android’s open architecture allows users to install apps from third-party stores or directly from the web through a process known as sideloading. Google has taken steps to warn users about the risks of sideloading and has implemented Google Play Protect, a built-in security feature that scans apps installed from any source for malicious behavior.
According to Google, Play Protect performed 125 billion app scans per day across Android devices in 2025, identifying threats on devices even when apps were not installed through the Play Store. This on-device scanning capability is particularly important in markets where third-party app stores are prevalent, such as parts of Southeast Asia and Africa. Google has also expanded real-time scanning capabilities that analyze app behavior after installation, catching threats that evade initial static analysis.
Developer Accountability and the SDK Problem
One of the more nuanced challenges Google faces involves software development kits, or SDKs, that are embedded within otherwise legitimate apps. SDKs are third-party code libraries that developers incorporate into their apps to add functionality such as advertising, analytics, or social media integration. In some cases, these SDKs collect user data in ways that violate Google’s policies, or they introduce vulnerabilities that can be exploited by attackers. The developer of the app may not even be aware that the SDK they are using is problematic.
Google said it worked with SDK providers throughout 2025 to address these issues, and it expanded its SDK transparency requirements. Developers are now required to disclose the SDKs used in their apps and to ensure that those SDKs comply with Google’s data safety policies. When an SDK is found to be in violation, Google notifies all developers using it and gives them a deadline to update or remove the offending code. Apps that fail to comply are removed from the store. This approach attempts to address the root cause of many policy violations rather than simply treating the symptoms on an app-by-app basis.
How Google’s Efforts Compare to Apple’s App Store
The scale of Google’s enforcement efforts inevitably draws comparisons to Apple’s App Store, which operates under a more restrictive model. Apple has historically touted its rigorous human review process as a key differentiator, arguing that its curated approach results in fewer malicious apps reaching users. However, Apple has faced its own challenges, including high-profile incidents where malicious apps evaded its review process. The fundamental difference lies in the sheer volume of submissions: Google’s Play Store receives far more app submissions than Apple’s App Store, in part because of Android’s larger global market share and the lower barriers to entry for developers.
Google’s reliance on AI to handle this volume is a practical necessity. The company has been transparent about the fact that no automated system is perfect, and it continues to invest in human review teams that handle the most complex and ambiguous cases. But the direction of travel is clear: AI is doing more of the heavy lifting each year, and Google is betting that continued improvements in machine learning will allow it to stay ahead of increasingly sophisticated threats.
Regulatory Pressure and the Road Ahead
Google’s security disclosures come at a time of increasing regulatory scrutiny over app store practices worldwide. The European Union’s Digital Markets Act, which took effect in 2024, has imposed new obligations on large platform operators, including requirements around transparency in content moderation and app review processes. In the United States, lawmakers have introduced legislation that would require app store operators to provide more detailed reporting on their security practices and the prevalence of malware on their platforms.
For Google, the annual security report serves a dual purpose: it is both a public relations exercise designed to reassure users and regulators that the company is taking security seriously, and a genuine accounting of the threats facing the Android platform. The 2025 figures suggest that the threat is not diminishing — if anything, it is growing in both volume and sophistication. Google’s AI systems are catching more malicious apps than ever, but the attackers are adapting just as quickly, creating a dynamic that shows no signs of stabilizing.
What This Means for Android Users and the Broader Industry
For the average Android user, Google’s 2025 report offers a mixed message. On one hand, the company’s AI-powered defenses are intercepting millions of threats before they reach users. On the other hand, the sheer volume of attempted attacks — and the acknowledgment that some malicious apps still slip through — is a reminder that no platform is entirely safe. Users are still advised to exercise caution when installing apps, to pay attention to permission requests, and to keep Google Play Protect enabled on their devices.
For the broader technology industry, Google’s experience highlights a growing consensus: AI is becoming indispensable for security operations at scale. The volume of threats facing major platforms has simply outpaced what human reviewers can handle, and machine learning offers the only viable path to keeping up. As TechCrunch reported, Google’s investment in AI-assisted security review is part of a larger trend across the industry, with other major platforms including Microsoft and Meta also expanding their use of automated systems to detect and remove harmful content and software. The question going forward is not whether AI will play a central role in platform security, but whether it can evolve fast enough to match the ingenuity of those who seek to circumvent it.