SquareX warns browser flaws now key threat to firms

Cybersecurity firm SquareX has warned that structural weaknesses in popular web browsers now present a major risk for enterprises, after publishing a year-long body of research into browser-based attacks.

The company said its Year of Browser Bugs programme uncovered serious flaws in browser extensions, AI-driven browsers and identity systems that sit inside the browser.

Researchers released 11 pieces of research over the past 12 months. They disclosed several zero-day vulnerabilities at security conferences including DEF CON, Black Hat, RSA and BSides.

SquareX said the work shows that the browser has become the key entry point for attackers against corporate systems. It also said traditional endpoint and network tools do not fully address these threats.

"When we launched YOBB earlier this year, SquareX has been seeing a lot of browser native vulnerabilities that are being used to exploit enterprises and users," said Vivek Ramachandran, Founder of SquareX. "These disclosures demonstrate that architectural limitations across browsers, extensions, and emerging AI technologies are putting organizations at risk in ways traditional security solutions simply cannot address. This demonstrates the importance of security teams to keep up with the modern way of working, whether it is securing AI or the browser itself."

SquareX described the browser as the new endpoint for the enterprise. Employees now use it as the main interface for SaaS applications and online data.

The firm said this shift has created a gap in defences. It also said attackers increasingly exploit advanced browser features and new AI browsing tools.

Extension flaws

One part of the research focused on browser extensions. These are small add-ons that users install in browsers.

SquareX found what it called fundamental flaws in how extensions operate and how they are monitored.

In January, its Browser Syncjacking work showed that a malicious extension could take over user profiles, the browser and devices with minimal permissions. It said this attack used standard browser synchronisation features.

In February, research into Polymorphic Extensions showed that attackers could closely imitate legitimate tools. The examples included password managers and cryptocurrency wallets.

The company said such extensions could steal credentials while appearing trustworthy.

In July, SquareX published work on what it described as architectural limitations in browser developer tools. It said these tools offer little visibility into how extensions behave.

The firm said this limits what end users and security teams can do when they inspect extension activity.

AI browser risks

SquareX also examined AI-powered browsers and browser-based AI agents. These tools are gaining use in enterprises.

The company said its own data showed that 79% of organisations now deploy agentic workflows. These workflows automate tasks through AI systems.

In June, research on Browser AI Agents concluded that such agents are trained to complete tasks rather than recognise threats. SquareX said this makes them more exposed to attacks than human workers.

In September, the firm detailed what it called architectural security vulnerabilities in AI browsers. It said attackers could abuse these designs to extract data, distribute malware and gain unauthorised access to SaaS applications.

In October, research on an AI Browser Sidebar Spoofing attack described how a malicious extension could create a perfect copy of an AI sidebar inside the browser. The fake sidebar displayed false instructions.

SquareX said this technique could lead users into phishing pages, malicious file downloads and possible device takeover.

In November, researchers disclosed an issue with a poorly documented MCP API in the Comet browser. SquareX said the API allowed embedded extensions to run arbitrary local commands without explicit user permission.

The company said this behaviour could include execution of known ransomware.

Data and identity

SquareX also examined attacks that target data and identity within the browser.

In April at BSides SF, its researchers presented work on Data Splicing Attacks. These techniques assemble sensitive information in the browser in ways that avoid inspection.

The company said the methods bypass all data loss protection products listed in Gartner's Magic Quadrant.

In March, research on Browser Native Ransomware outlined how attackers could run a ransomware-style attack from within the browser. The attack did not rely on traditional local files or visible processes.

SquareX said such attacks instead abuse identity mechanisms and session data that reside in the browser.

In May, the Fullscreen BitM (browser-in-the-middle) attack showed how one click could place a user in a full-screen, attacker-controlled browser window. The user believed they were interacting with a legitimate site.

The company said this setup allowed full monitoring of credentials that the victim entered.

At DEF CON 33 in August, SquareX delivered a talk titled "Passkeys Pwned: Turning WebAuthn Against Itself". The session covered a passkey implementation flaw.

The firm said the flaw allowed a malicious script or extension to gain unauthorised access to enterprise SaaS applications and resources.

Ongoing research

SquareX has compiled its findings into a Year of Browser Bugs report. The document brings together the different research strands from the past 12 months.

The company said it would continue to focus on browser detection and response as an emerging security category. It plans further work on threats from AI agents, malicious extensions and browser-based identity attacks.

SquareX said it expects attackers to keep targeting the browser as the most common entry point into modern enterprises.