Google's browser has taken a significant leap forward by integrating artificial intelligence models capable of recognizing and stopping scams in real time while we browse. This new layer of protection focuses on very common trapssuch as fake tech support alerts, deceptive notifications, and sites that try to lock the screen to force a phone call to a fraudulent number. The goal is to stop the deception before the user falls for it.and do it without sending your data to the cloud.
The key is that the analysis runs locally on your computer or mobile device, privately and quickly. This allows Chrome to warn you about suspicious pages even if they are new and you've never seen them before. This protection is integrated with Safe Browsing in its Enhanced Protection mode.and is progressively spreading to more devices and types of scams.
How Chrome's AI anti-fraud works
Google has introduced a lightweight model called Gemini Nano that analyzes page content as it loads. It works on the device itself, reducing latency and, most importantly, preserving privacy because it doesn't need to send content to external servers. The system evaluates visual and textual signals to decide if something smells like fraud.
Among the problematic signs you can identify are alleged full-screen system errors, beeps or loud, alarming sounds that try to scare you, system window spoofing, and imperative instructions such as "call now" or "download this update". You might even notice technical tactics such as using the keyboard lock API to make it harder for you to close the tab.
When the model detects these patterns, it alerts Safe Browsing to issue a warning and block the page if necessary. Execution is asynchronous, with careful management of GPU and memory, so your browser won't lag mid-task. The idea is to intervene only when necessary and without penalizing the user experience.
This AI-powered approach is especially useful against scams that change their appearance in a matter of minutes. Many malicious sites appear and disappear in less than ten minutes, so traditional blacklists are too late. The evaluation on the device allows for on-the-fly reaction.even when faced with traps that were not listed.
Technical support scams: the priority target
Scams that pretend to be technical support are one of the most common traps used by fraudsters. Their script is well-known: they lock the screen with an alarming message, play an annoying sound, display a fake phone number, and try to get you to call to "fix" a non-existent problem. The end result is usually the remote control of the equipment. or the installation of malicious software.
Chrome's AI recognizes these patterns and warns you before you take the next step. If it detects that the site replicates this scenario, it displays a visible warning and suggests closing the page. This protection is already being applied in the desktop version for those who activate Enhanced Protection, and Google plans to expand its reach.
In addition to fake messages and overlays, the system is able to detect social engineering “tricks”: fake logos of well-known companies, timed instructions to create a sense of urgency, or pages that simulate system dialog boxes. Contextual analysis helps reduce false positives because it understands not only the words, but how and where they appear.
Deceptive web notifications on Android
The risk doesn't end when you close a tab. Many sites try to sneak into your notifications to continue bombarding you with ads and malicious links. cut that channelChrome on Android incorporates machine learning models that work offline and analyze whether a notification could be deceptive. If there are indications of manipulationYou will see a notice with the option to unsubscribe or keep them if you believe it was a mistake.
This feature also comes into play when a website asks for permission to send notifications: if the site looks suspicious, the system alerts you before you accept. According to Google, this approach has significantly reduced complaints and accidental clicks on malicious ads. helping to curb spam and fraud From the phone.
An AI that adapts to the moment and on a large scale
The same early detection philosophy applies to the Search Engine. The company reports that its AI systems have detected and blocked hundreds of millions of fraudulent results every day. Improvements in classifiers and models They have increased the detection of deceptive sites by 20 times compared to previous methods.
This has allowed authorities to tackle coordinated campaigns, such as fake airline customer service lines, where their presence in search results has been reduced by more than 80%. Websites that mimic official procedures or visa applications have also decreased by more than 70%. The combination of massive text analysis and behavioral cues help anticipate new tactics.
Applied to the browser, this continuous learning means that even if a malicious website only survives for a few minutes, there is still time to respond. Thanks to local analysis and contextual understanding, More frauds are detected with fewer false positivesavoiding unnecessary alarms that erode user confidence.
What other traps will it detect and how will it be deployed?
Google has already explained that it will expand its scope beyond tech support scams. The roadmap includes common frauds such as fake package tracking or threats related to alleged unpaid tolls. The system will be deployed in phases. both on desktop and on Android.
On computers, Gemini Nano is integrated into Chrome 137 to perform real-time analysis and trigger warnings when it detects suspicious behavior on complex pages. On Android, notification protection and local content evaluation strengthen the barrier directly from the phone. The priority is to cover the most common attack vectors without depending on external connections.
The architecture is designed to activate only when there are risk signals, run in the background without blocking navigation, and manage device resources efficiently. Privacy and efficiency go hand in hand, with the added benefit of reacting to unprecedented scams.
Beyond anti-fraud: features with Gemini in Chrome
The integration of AI into the browser isn't limited to security. Chrome is incorporating assistance capabilities to summarize pages, answer questions, and clarify concepts, even combining information from multiple tabs. They are also working on an intelligent history feature. capable of retrieving visited content based on a question in natural language.
In the realm of everyday security, Gemini helps detect fake virus notifications, deceptive sweepstakes, and unauthorized permission requests. Improvements have been added to the password management with warnings of compromised credentials and the ability to change them in a guided way on compatible sites. The idea is to bring useful automation. without losing sight of protection.
Android vs. iOS: Differences in protection against scams
Google boasts that its proactive approach, especially with the Pixel line, drastically reduces exposure to fraud. According to internal data and a YouGov survey across several countries, Pixel users report up to 96% fewer malicious messages than iPhone users. In the Android add-onUsers receive 58% fewer fraudulent messages than on iOS.
The company attributes this difference to how AI is implemented: Android integrates it across calls, notifications, the system, and apps, while Apple concentrates it in more limited areas such as Messages and the App Store. The data from Spain also reflects Android's dominance, with a 75,6% market share compared to iOS's 24,1%. With more devices under its umbrella, improvements scale rapidly.
Apple, for its part, has adopted features such as call filtering in iOS 26, a feature that was already available on Android in the Phone app. Although the approaches converge, Google emphasizes its advantage due to accumulated experience and the integration of AI at the platform level.
Performance and consumption: what you can expect
Chrome's AI-powered fraud protection is designed to interfere only when necessary: ​​it runs asynchronously and only when it finds signals, with special care taken with GPU and memory usage. That's the plan to avoid disrupting navigation.Even so, it is important to distinguish these functions from other AI utilities that may demand more resources.
In tests with features such as page summaries read aloud and searches in a so-called "AI Mode", average RAM increases of 42% were observed with three active tabs, and CPU peaks of 78% when generating complex summaries (compared to 15%-20% for normal browsing). This does not imply that the anti-fraud module consumes as much.But it's a useful reminder that AI has a computational cost that developers must optimize.
On modest devices, adjusting the use of advanced AI features can make a difference. More granular options to limit their resource consumption per tab or extension would be desirable. preventing background tasks penalize overall performance when we don't need them.
How to enable Enhanced Protection in Chrome
To take advantage of instant alerts and AI-powered analysis, turn on Enhanced Safe Browsing Protection in Chrome. It's a quick setting that significantly strengthens your security against phishing, malware, and new scams. The steps are simple and it takes less than a minute.
- Open Chrome and go to Settings.
- Go to Privacy and security.
- Select Safe Browsing.
- Choose Enhanced Protection.
With this option enabled, Chrome will use its models and signals to warn you about dangerous sites and downloads, and on desktop it will activate the additional layer powered by Gemini Nano when needed. The data collection for security improvement is done anonymously. and with safeguards designed by Google.
Good practices to avoid falling into the trap
Chrome's AI has your back, but the first line of defense is always common sense. Avoid clicking on ads that threaten you or offer free prizes, scrutinize suspicious URLs, and don't call numbers that appear in pop-ups. Keep Chrome updated to its latest version and use official sources for downloads and support.
If a website requests notification permissions without reason, deny them and move on. Blocking abusive pop-ups can help you avoid many of these traps. And if something smells fishy to youClose the tab; it's better to lose a minute than to expose your data.
Privacy: what you should never ask an AI to do
The arrival of assistants like Gemini in more Google products opens up useful possibilities, but it's important to remember that not everything should be shared with AI. There are categories of information you should never enter, no matter how tempting it is to ask for help with them. Protect your identity and assets It also depends on what you choose not to reveal.
- IIP extremely sensitive (ID card with processing number, passport, CUIT/CUIL, driver's license): can be used for impersonation and fraudulent procedures.
- Credentials and passwordsThey are not key managers; use a password manager and enable two-step verification.
- Complete financial data (card number, CVV, bank account): direct risk of fraud and unauthorized transactions.
- Private medical information (diagnoses, genetics, treatments): can lead to discrimination or blackmail if leaked.
- Biometric data (fingerprint, face, iris): are irreplaceable; their exposure compromises your identity for life.
- Trade secrets or intellectual propertyDisclosing sensitive information can cause you to lose legal protection and competitive advantage.
Golden rule: if you wouldn't share that data via unencrypted email, don't upload it to a general-purpose AI system either. In case of error, act quickly: change passwords, block cards and notify security officers.
Results and evolution of the ecosystem
The anti-fraud offensive isn't limited to Chrome. In Search, AI advancements have enabled the blocking of 20 times more deceptive pages and mitigated campaigns like fake airline "help desks." On Android, protections have been added to Messages, along with features to detect suspicious calls, including tactics where scammers try to trick you into opening your banking app mid-conversation. All of this converges in Android 16, which strengthens default security with more preventative measures.
This multi-layered approach explains differences in exposure between platforms, and Google argues that its “top-down” AI integration allows it to anticipate and neutralize fraud more effectively. The central idea is the same on all fronts: less reliance on static lists, more real-time contextual analysis.
As these protections extend to more countries and devices, we will see the catalog of detected frauds expand: from supposed "pending" toll payments to fake package tracking attempts to steal credentials. The phased deployment It will help measure impact and adjust models without sacrificing performance.
Chrome's combination of local AI to detect scams on the fly, proactive warnings on Android against deceptive notifications, and constant cleaning of the Search engine with large-scale models creates a robust lock for most users. With Enhanced Protection active and a little common senseYou'll have a browser that not only reacts, but anticipates the most common traps, keeping your data safe and your browsing experience smooth.
