AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Chrome warns you about suspicious websites now – how to turn on Google’s new scam protection

Collected Oct 21, 2026

Google announced on Tuesday three new anti-scam measures for Chrome, applicable to both desktop and mobile, in a blog post describing work aimed at keeping users from falling for phishing, fake support scams and social engineering. The headline feature is a second tier of warning: Chrome already blocks and flags sites known to be malicious, and it will now also flag sites judged merely suspicious.

The suspicious-site warning exists because the browser can compare a page against known deceptive sites and spot similarity to pages tied to phishing, malware and scams. When the warning appears, it tells the user the site does not seem right, offers a Learn More link for detail, and gives two paths: leave via a Return to safety button, or mark the site as safe and continue if the user is confident it is legitimate.

Google judges suspicion through a mix of site comparison and a Gemini model that examines page content, the brand represented, apparent intent, domain reputation and other signals to produce a scam likelihood score. The two warning tiers are deliberately styled differently: a dangerous-site alert fills the screen red and presses the user to leave, while a suspicious-site alert is quieter and leaves the door open. Google said the new warnings differ by format, iconography, color and text so users can decide quickly and confidently, and that tailoring the interface this way lets it safely issue a higher volume of protective warnings.

That design choice is the interesting part. Aggressive warnings train people to click through them, so carving out a softer tier lets Google widen coverage without crying wolf on every borderline page. The trade-off is that the softer warning depends entirely on the accuracy of a likelihood score, and a false negative keeps a user on a scam site. Allowing users to override the suspicious verdict is reasonable for a judgment call, but it also hands scammers a plausible-sounding page to point at when they coach victims to dismiss warnings.

Second, Google widened the AI-driven scam detection it brought to Chrome last year for tech support scams, in which a malicious page tries to convince a visitor that malware has infected the machine and then sells security software or seeks remote access. The same approach, using a server-side Gemini model with more advanced capabilities than before, now covers a broader set of scams including fake antivirus alerts and phony browser update prompts. The model weighs page content, domain reputation and other signals to reach a conclusion.

Google addressed the obvious privacy concern: the data sent to its back-end servers is kept to a minimum and transmitted only when a site trips high levels of scam signals. Page content sent from the device is not logged, not remembered and not used to train the AI, the company said. This is the architecture to watch. Server-side models can be stronger and updated faster than something shipped in the browser, but every request is a disclosure of page context to Google, so the stated minimization and no-logging commitments are what make the feature acceptable to privacy-conscious users.

Third, Chrome now defends against ClickFix, a social engineering trick that persuades a victim to copy and manually run malicious commands on their own machine, typically framed as fixing a problem that does not exist. Because the user performs the steps, conventional defenses have had little to catch. In the new behavior, Chrome notices when a user has been induced to copy suspicious commands, wipes personal data from the clipboard before a paste can happen, analyzes the copied content and shows a warning if it looks dangerous. This targets the exact gap attackers exploited: the payload travels through the user's own hands rather than through a download. In practice it will likely be most effective against scripted, copy-paste command lures and most vulnerable to a determined user who retypes the command from a page's instructions.

All of this is gated behind Enhanced Protection, the strongest tier of Chrome's Safe Browsing. On desktop or mobile, users open Settings, choose Privacy and security, then Security on desktop or Safe Browsing on mobile, and select Enhanced Protection. Google also recommends turning on Enhanced protection for the Google account itself: sign in on the web, open Security & sign-in, scroll to Enhanced Safe Browsing and switch it on. Google framed the features as continuous real-time defenses and said turning on Enhanced Protection in Chrome and on the account is the way to receive them first. The practical consequence is that users on the default protection level will not see these warnings, so the benefit scales with how many people opt into the stricter mode.

Why it matters: product and security teams building on the web should expect Chrome to serve softer, opinionated warnings on pages it deems similar to known scam infrastructure, which could affect conversion funnels, lookalike marketing pages and aggressive interstitial flows. Because the heavier checks run server-side only when a site trips scam signals, sites that resemble fakes may face both user-facing friction and evaluation by a Gemini model whose thresholds Google has not published; treat that opacity as a reason to monitor how your domains are classified. For anyone supporting end users, the ClickFix clipboard intervention is the most concrete change, since it breaks an attack chain that previously relied on the victim's own actions, though it is inference that determined users will find ways around it.

Read at ZDNet · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Google’s browser will flash two types of warnings based on whether the site is deemed dangerous or just suspicious.