![\"Chrome](\"https:\/\/www.newsbeep.com\/ca\/wp-content\/uploads\/2025\/08\/1755604867_793_960x0.jpg\")
<\/p>\n<\/p>\n
Most users are at risk.<\/p>\n
NurPhoto via Getty Images<\/p>\n
Here we go again. While Google\u2019s procession of critical security fixes and zero-day warnings<\/a> makes headlines, the bigger threat to its 3 billion users is hiding undercover.<\/p>\n There\u2019s \u201ca new class of exploit\u201d targeting \u201cthe top 5 GenAI tools\u201d through a previously overlooked vector: the browser extension.” LayerX<\/a> says this means \u201cbillions could be affected,\u201d with 99% of users \u201cpotentially exposed to this attack vector.”<\/p>\n Just 3 weeks ago, SquareX<\/a> warned that \u201cmillions of users have data stolen,\u201d also by malicious or hijacked browser extensions, with security tools having no runtime visibility to \u201cprotect users against the rising threat vector.\u201d<\/p>\n ForbesApple Encryption Fight\u2014U.S. Forces U.K. To Drop \u2018Back Door\u2019By Zak Doffman<\/a><\/p>\n You can see the theme here. The primary threat is to Windows PCs, which means Google Chrome which dominates and essentially controls the browser market \u2014 at least until the new genre of AI browsers gets any kind of foothold.<\/p>\n And it\u2019s those same AI platforms that are behind LayerX\u2019s \u201cMan-in-the-Prompt Attack\u201d warning. The team says its research \u201cshows that any browser extension, even without any special permissions, can access the prompts of both commercial and internal LLMs and inject them with prompts to steal data, exfiltrate it, and cover their tracks.\u201d<\/p>\n They warn AI assistants will become \u201c\u2019hacking copilots\u2019 to steal sensitive corporate information.” When you use an AI assistant, \u201cthe prompt input field is typically part of the page\u2019s Document Object Model (DOM). This means any browser extension with scripting access to the DOM can read from, or write to, the AI prompt directly.\u201d<\/p>\n Every LLM, AI Application is Affected<\/p>\n LayerX<\/p>\n We are seeing more noise about direct and indirect prompt injection attacks, where your AI assistant sees what you don\u2019t, as hidden instructions buried within a document or even an online search result trick the AI assistant into doing something it shouldn\u2019t, compromising you, your data and your workplace.<\/p>\n \u201cBecause of this tight integration between AI tools and browsers,\u201d LayerX says, \u201cLLMs inherit much of the browser\u2019s risk surface.\u201d This is made worse because your browser extension usually operates with your security credentials, as if it\u2019s you taking its actions. That can include signing into websites, even if the sign-in is a malicious fake.<\/p>\n It\u2019s the same kind of risk<\/a> that prompted Google<\/a> to warn that \u201cwith the rapid adoption of generative AI, a new wave of threats is emerging across the industry with the aim of manipulating the AI systems themselves.\u201d<\/p>\n