A new security flaw in agentic web browsers, such as OpenAI ChatGPT Atlas, has been identified by cybersecurity experts. It leaves underlying artificial intelligence (AI) models vulnerable to context poisoning attacks.
A malicious actor can create webpages that provide different content to browsers and ChatGPT and Perplexity-powered AI crawlers as part of the assault developed by the AI security firm SPLX. The method is known as AI-targeted cloaking.
The strategy is a variant of search engine cloaking, which is the practice of showing visitors one version of a webpage while giving search engine crawlers a separate version in an attempt to manipulate search engine rankings.
In this instance, the only distinction is that attackers employ a simple user agent check to optimize for AI crawlers from different providers, which results in content delivery manipulation.
“Any content that is provided to these systems becomes ground truth in AI Overviews, summaries, or autonomous reasoning because they rely on direct retrieval,” stated security experts Ivan Vlahov and Bastien Eymery read more about New AI-Targeted Cloaking Attack Tricks AI Crawlers Into Citing Fake Info as Verified Facts.
Get up to date on the latest cybersecurity news and enhance your knowledge of cybersecurity with our thorough coverage of the dangers, breaches, and solutions.