AI browsers raise new fraud risks for GCC users
As AI–powered browsers transition from experiment to everyday tool, a new class of digital risk is emerging beneath their sleek interfaces.
Platforms such as Arc Search, Atlas, and Perplexity’s Comet now read pages, summarize content, and even complete purchases on behalf of users, blending search, chat, and shopping in a single AI layer. But, security researchers warn that these “smart” browsers can quietly expand the attack surface for scams and data theft.
Unlike traditional browsers, AI browsers continuously send page content and user prompts to cloud models, raising the risk of sensitive data exposure and “shadow AI” usage that bypasses corporate controls. Network security firms note that legacy secure web gateways often cannot detect or control these new traffic patterns, making it harder for security teams to track where data flows.
At the same time, vulnerabilities in AI browser extensions are under scrutiny. A recent dispute between security researchers and Perplexity over an alleged critical issue in the Comet browser involved a hidden API that could be abused under certain conditions. This highlighted how quickly a misconfigured AI agent could enable device compromise, even as the company disputes key details.
The fraud landscape is evolving just as fast. Global data show that over 82% of phishing emails are now generated with AI, and GenAI-enabled scams grew 456% between May 2024 and April 2025. In the Gulf, deepfake-enabled attacks have already targeted banks and enterprises, including a $35 million CEO fraud case linked to a United Arab Emirates executive, while deepfake volumes across the Middle East are rising sharply.
For the Gulf Cooperation Council (GCC), where governments are doubling down on generative AI adoption and cybercriminal forums increasingly target the region, AI browsers can boost productivity, but they must be treated as high-risk applications. Enterprises need explicit policies, monitoring, and user training; consumers should verify sources, avoid sharing sensitive data with AI pages, and treat AI-generated “support” or payment links with caution.
