![\"Perplexity](\"https:\/\/www.newsbeep.com\/ca\/wp-content\/uploads\/2025\/10\/Perplexity-Comet-AI-browser-discussing-its-security-problems.jpg\")
<\/p>\nComet is vulnerable to LLM exploits<\/p>\n
Large language models are vulnerable to something called \u201cprompt injections,\u201d which can happen whenever the LLM is made to process and interpret text. The problem is that when text is fed into an LLM, it isn\u2019t always clear whether that text is from you or elsewhere.<\/p>\n
For example, an LLM might process and analyze the source code of a web page to perform some kind of task. But the source code of a web page could potentially include hidden prompt instructions designed to hijack an AI that\u2019s analyzing the source code. The LLM can\u2019t distinguish the hidden prompt in the source code from the source code itself.<\/p>\n
In other words, that hidden prompt was injected into the source code, and the AI will be none the wiser. Hence, prompt injection attack.<\/p>\n
<\/p>\n
Chris Hoffman \/ Foundry<\/p>\n
Security researchers at Guardio found that Comet was vulnerable to attacks like this<\/a>, and that it could be tricked into falling for phishing scams while online shopping. Security researchers from Brave also found that Comet was vulnerable to indirect prompt injection attacks<\/a>. Here\u2019s the wildest part from Brave\u2019s blog post:<\/p>\n \u201cThe vulnerability we\u2019re discussing in this post lies in how Comet processes webpage content: when users ask it to \u2018Summarize this webpage,\u2019 Comet feeds a part of the webpage directly to its LLM without distinguishing between the user\u2019s instructions and untrusted content from the webpage. This allows attackers to embed indirect prompt injection payloads that the AI will execute as commands.\u201d<\/p>\n Did you catch that? It\u2019s not that Comet\u2019s protections against prompt injection were bypassed, but rather that Comet (in its initial release) didn\u2019t even have prompt injection protections that tried to distinguish between trusted user instructions and untrusted web page data sent to the AI model (at least with the summarization function).<\/p>\n This sort of thing is a known problem with large language models. While Comet now has better protections against this, it\u2019s unclear how good those safeguards are. Comet hasn\u2019t been properly battle-tested.<\/p>\n Other agentic AI browsing solutions\u2014like ChatGPT\u2019s agent mode\u2014interact with websites by loading those websites in their own browser in the cloud apart from your data. Even when those LLMs are exploited by prompt injection attacks, at least the damage is somewhat limited.<\/p>\n But when the AI has access to everything in your browser\u2014as is the case with Perplexity\u2019s Comet\u2014the risk goes up by quite a bit. From what I can tell, it seems like Perplexity is \u201cmoving fast and breaking things\u201d while competitors are at least paying attention to security before launching.<\/p>\n","protected":false},"excerpt":{"rendered":"Comet is vulnerable to LLM exploits Large language models are vulnerable to something called \u201cprompt injections,\u201d which can…\n","protected":false},"author":2,"featured_media":191393,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[62,276,277,49,48,61],"class_list":["post-191392","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-artificial-intelligence","tag-artificialintelligence","tag-ca","tag-canada","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts\/191392","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/comments?post=191392"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts\/191392\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/media\/191393"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/media?parent=191392"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/categories?post=191392"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/tags?post=191392"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}