Summary
- Chrome now distinguishes suspicious websites from confirmed dangerous destinations through separate warning presentations.
- A server-side Gemini model assesses potential scams alongside domain reputation and other Safe Browsing signals.
- New ClickFix detection examines suspicious clipboard commands and warns when Safe Browsing identifies a dangerous payload.
Google Chrome has introduced additional protections against deceptive websites and ClickFix attacks, expanding the browser’s Enhanced Protection mode with new warning categories, AI-assisted scam analysis and checks on suspicious commands copied from web pages.
The capabilities, detailed on 20 October, are designed to identify threats that attempt to persuade users into performing harmful actions rather than simply downloading an obviously malicious file.
Google has introduced separate warnings for websites considered suspicious and those classified as dangerous. It has also moved its Gemini-assisted scam analysis to a server-side model, while adding a detection process for potentially malicious commands placed on a user’s clipboard.
The changes apply to Chrome’s Enhanced Protection mode, which provides additional security checks beyond those available through the browser’s Standard Protection setting.
Warnings distinguish suspicion from confirmed danger
Chrome’s Safe Browsing system has traditionally displayed prominent warnings when a website is known to present a significant phishing or malware risk.
The latest changes introduce a separate category for sites displaying deceptive characteristics without meeting the higher-confidence threshold associated with known dangerous destinations.
Suspicious-site warnings use different colours, icons, wording and layouts to distinguish a possible threat from a website classified as dangerous.
Chrome will continue to use its familiar full-page red warning for sites receiving a high-confidence dangerous verdict.
The additional category allows Google to warn about more pages while communicating that the available evidence does not justify treating each one as a confirmed malicious destination.
This approach reflects a challenge in detecting social engineering. Fraudulent sites may imitate legitimate businesses, browser notifications or security services without hosting an immediately recognisable malicious executable.
A warning based on deceptive patterns can therefore provide an intervention before a page has been associated with a fully established attack campaign.
However, a suspicious verdict remains different from proof that a particular site is malicious. Google says the distinction is intended to help people interpret the level of risk reflected in the browser’s warning.
Gemini analyses deceptive page characteristics
Google has also extended its use of Gemini models within the Safe Browsing detection process.
An earlier on-device implementation supported detection of technical support scams on compatible devices. The company has now moved to a more capable server-side model to provide broader scam assessment for users with Enhanced Protection enabled.
The model examines characteristics of potentially deceptive pages, including the brand they claim to represent, their apparent purpose and indicators associated with fraudulent behaviour.
Its assessment is combined with domain reputation and other security signals before Safe Browsing produces the final verdict.
That design means the AI model’s assessment is not described as the sole basis for a warning. It supplies additional information to an existing security decision process.
Google says page text is sent to the server for this analysis only when the site exhibits high-confidence scam indicators.
The company also states that transmitted page content is processed ephemerally, without being logged, retained or used to train its models.
The handling restrictions distinguish this feature from unrestricted remote analysis of every page a person visits.
Nevertheless, the server-side approach involves different processing from the earlier local model, making the conditions under which content leaves the device relevant to the feature’s privacy assessment.
The expanded analysis covers threats such as fraudulent antivirus messages and deceptive browser update requests, extending the scope beyond the original technical support scam focus.
ClickFix moves malicious activity outside the browser
The new command detection capability addresses ClickFix, a social engineering technique in which a webpage persuades a person to execute instructions on their computer under the pretence of repairing an error or completing a verification step.
A fraudulent CAPTCHA or browser warning may instruct the visitor to copy text and run it through a terminal or command interface.
The command actually placed on the clipboard can differ from the text displayed on the page, reducing the chance that its purpose is apparent.
Attackers often use legitimate operating system utilities because they are already installed and capable of executing commands or retrieving additional software.
Windows PowerShell and the command processor are among the tools that can be misused in such attacks.
By persuading a visitor to run a command outside the browser, the attacker attempts to bypass protections designed to examine downloads initiated through normal browser processes.
The malicious activity is therefore not necessarily caused by exploiting a browser vulnerability. It may depend on persuading the user to authorise an operation the browser could not perform directly.
That makes the transition from webpage instructions to clipboard content an additional point at which suspicious behaviour can be assessed.
Clipboard analysis adds a new intervention point
Google says Chrome can now detect when an Enhanced Protection user copies a command resembling a potential ClickFix payload.
When this occurs, the clipboard content is scrubbed of personally identifiable information before being submitted to Safe Browsing for analysis.
If the service returns a dangerous verdict, Chrome displays a warning.
The process is designed to minimise disruption to legitimate development and administration work, where copying commands from documentation is a normal activity.
Not every copied command is therefore treated as malicious, and the existence of a warning does not establish that the command has already been executed.
The mechanism is also distinct from conventional malware scanning. Instead of examining a file downloaded through the browser, it evaluates instructions that a website may be trying to persuade the user to execute elsewhere.
This does not mean Chrome can prevent every malicious command from being run. The published description concerns detection and warning behaviour within Enhanced Protection, not universal control over terminal applications or operating system processes.
Effectiveness claims require careful interpretation
Google says more than one billion Chrome users had enabled Enhanced Protection by 2025 and describes the mode as offering substantially stronger protection than Standard Protection.
Its latest published comparison uses the average number of security warnings shown to users under the two settings.
That measurement indicates a difference in the frequency of protective interventions. It does not independently demonstrate that a corresponding number of successful compromises were prevented.
The distinction is relevant when evaluating the effectiveness of new browser security functions, particularly those involving probabilistic assessments of suspicious content.
Expanded warning coverage may identify additional threats, but it also requires careful handling of uncertain results and legitimate activities that resemble attack patterns.
Google’s two-tier site warnings and command analysis are intended to preserve those distinctions rather than classify every unusual page or clipboard operation as an established attack.
The October release adds these capabilities to Chrome’s Enhanced Protection offering. Its product changes are separate from individual threat investigations, including campaigns in which attackers have used sponsored search results and counterfeit verification pages to distribute malware.
The company has not published a complete count of attacks prevented specifically by the new ClickFix feature. Its confirmed development is the addition of suspicious-site classification, broader Gemini-assisted scam assessment and clipboard command warnings to the browser’s Enhanced Protection defences.
