00:00What if just visiting a web page could let hackers into your email or bank account?
00:04That nearly happened with Perplexity AI's new browser.
00:08Perplexity AI's Comet browser had a dangerous flaw.
00:11Its built-in AI assistant could be hijacked by hidden commands inside normal web pages.
00:16Researchers at Brave discovered the trick.
00:18And it's shockingly simple.
00:20A Reddit post looked innocent.
00:22But behind a spoiler tag was a hidden message.
00:25When a user clicked Summarize this page,
00:28the AI assistant read everything, including the secret commands.
00:32That hidden text told the assistant to fetch emails,
00:35grab a one-time password,
00:37and give a hacker full control over the account.
00:40No fancy hacking skills needed.
00:42Just clever wording.
00:43Brave warned that this type of attack bypasses traditional web security
00:47because the AI acts with full user privileges.
00:50It can expose emails, bank accounts, even company systems.
00:54Perplexity says it fixed the issue.
00:57But Brave isn't convinced.
00:59They've re-reported the problem.
01:00As AI becomes part of everyday browsing.
01:03Experts warn, without proper guardrails.
01:06AI could become the Internet's next biggest security threat.
01:10Security and AI must go hand in hand.
01:12Or users could pay the price.
01:14Are there any more problems.
01:15Noibus?
01:16Be able to knock others out.
01:17Flew into the country or người是啊,
01:22Amen.
01:33Read a 58% bite at a way.
01:34It is honestly for the explanation through writing.
01:35readings
01:36The fl встрет in the logic.
01:37Ah, really?
Comments