The Perplexity AI browser has serious security vulnerabilities, and user data may be at risk of leakage.

On August 26, according to security company Brave Software, Perplexity AI has a browser named Comet that contains a serious security vulnerability, allowing attackers to embed hidden commands in web pages to trick the AI assistant into leaking users' private data, including emails and code.

Brave researchers have demonstrated that when users ask Comet to summarize Reddit pages containing hidden commands, the AI assistant executes those commands. Although Perplexity claims the issue was fixed before being discovered and states that no user data was leaked, Brave insists that the vulnerability could still be exploited weeks after the fix, warning that Comet's design architecture makes it susceptible to further attacks.

View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • 6
  • Repost
  • Share
Comment
0/400
FunGibleTomvip
· 12h ago
Why can't the security vulnerabilities be fixed? Ridiculous.
View OriginalReply0
GasWastervip
· 18h ago
another L1 security fail... just like my failed txs during peak gas times smh
Reply0
TideRecedervip
· 18h ago
This product is doomed.
View OriginalReply0
rugdoc.ethvip
· 18h ago
I said earlier that AI is just a paper tiger now, and encryption too?
View OriginalReply0
GateUser-40edb63bvip
· 18h ago
Tsk tsk, caught red-handed and still refusing to admit it.
View OriginalReply0
GhostWalletSleuthvip
· 19h ago
Hehe, ChatGPT can dodge this, but this one can't.
View OriginalReply0
Trade Crypto Anywhere Anytime
qrCode
Scan to download Gate App
Community
English
  • 简体中文
  • English
  • Tiếng Việt
  • 繁體中文
  • Español
  • Русский
  • Français (Afrique)
  • Português (Portugal)
  • Bahasa Indonesia
  • 日本語
  • بالعربية
  • Українська
  • Português (Brasil)