Chrome adds AI checks for scams that trick users into running commands

Google is expanding Chrome’s Enhanced Protection mode with server-side Gemini analysis, graduated warnings and clipboard checks aimed at fake updates, bogus antivirus pages and ClickFix attacks.

Google is tightening Chrome’s anti-scam defences with a set of AI-driven checks that go beyond traditional blocklists. According to Google’s announcement on 20 October 2026, the browser now relies on a stronger server-side Gemini model to assess suspicious webpages, adds a lighter warning for borderline sites and begins checking commands copied during ClickFix attacks. The changes sit inside Safe Browsing’s Enhanced Protection mode, which is Google’s stricter, opt-in security setting. Android Central and TechPulse both reported the same broad shift in warning behaviour, while several security outlets noted that the update is aimed squarely at scams that pressure users into acting against their own interests.

The practical logic is straightforward. Criminals have moved away from simple fake login pages and towards lures that encourage people to install bogus antivirus software, accept a fraudulent browser update or paste a command into PowerShell or Command Prompt. Chrome’s new protections are designed to catch those patterns earlier. Instead of only checking whether a URL appears on a known-bad list, the browser is now trying to judge what a page is doing, how it is presenting itself and whether it is trying to push the user into risky action.

Google’s earlier work in this area used an on-device Gemini Nano model for some scam detection on supported desktop devices. The latest update shifts that judgement to a more capable server-side Gemini model for Enhanced Protection users. The move matters because a cloud-based model can weigh more signals at once and be updated centrally, which is useful when attackers constantly change domains, wording and visual design. Multiple reports said Google is using that model to widen coverage beyond older tech-support scam detection and into fake antivirus pages and fraudulent browser-update prompts.

The new warning system also introduces more nuance. Instead of treating every suspicious page as either safe or dangerous, Chrome will now show a lower-severity “This site doesn’t seem right” warning for pages that look deceptive but do not yet meet the threshold for a full block. Android Central described the high-severity response as a full-screen “Dangerous site” alert, while other reports said the middle tier is intended to fill the gap between no warning at all and an outright block. That distinction matters: a softer warning can alert users to risk without claiming certainty, but it also raises the possibility of warning fatigue if it appears too often.

ClickFix is the most notable addition because it targets a scam technique that largely bypasses traditional download scanning. In a ClickFix attack, the victim is tricked into copying and running a command themselves, usually under the pretence of fixing an error or completing a verification step. Chrome’s new defence inspects suspicious copied commands before they are pasted and run. Cyberstack reported that the clipboard content is scrubbed of personal information before analysis, and Google says a warning appears if the command is judged dangerous. The design is important because it intercepts the attack at the one point the scammer cannot avoid: the command has to reach the clipboard somehow.

These features are all tied to Enhanced Protection, not to Chrome’s default Standard Protection mode. That distinction is central to understanding the update. Google says Enhanced Protection users are ten times safer from phishing, social engineering, notification abuse and other scams than Standard Protection users, but the company has not published the methodology behind that claim in the material reviewed here. USA-24 and other outlets also reported that the new Gemini-based detection and ClickFix monitoring are opt-in, which means the reach of the update depends on how many users switch the setting on.

Privacy is part of the trade-off. Google says suspicious page content processed off-device is handled ephemerally and is not logged, retained or used to train models. For ClickFix, the company says personal information is removed from the copied command before analysis. Even so, the shift from local checks to server-side judgement means users are sharing more data in exchange for stronger detection. That is consistent with the logic of Enhanced Protection, but it also explains why some users and organisations may prefer to keep a stricter boundary around browser telemetry.

For everyday users, the advice remains simple. Chrome’s AI can reduce exposure to deceptive pages, but it will not eliminate the need for caution. Users should avoid running commands copied from websites, ignore pop-ups claiming a device is infected and treat urgent prompts to install software with suspicion. For administrators, the update is a reminder to review Chrome’s protection level across managed fleets and decide whether the stronger mode fits the organisation’s privacy and compliance requirements. In practice, browser-level AI works best as a second line of defence, not a replacement for careful behaviour and endpoint security.