Close Menu
Technology Mag

    Subscribe to Updates

    Get the latest creative news from FooBar about art, design and business.

    What's Hot
    Apple’s cheap laptop looks like a winner

    Apple’s cheap laptop looks like a winner

    March 7, 2026
    The Corvette ZR1X hybrid can outpace million-dollar sports cars for a fraction of the cost

    The Corvette ZR1X hybrid can outpace million-dollar sports cars for a fraction of the cost

    March 7, 2026
    DJI will pay K to the man who accidentally hacked 7,000 Romo robovacs

    DJI will pay $30K to the man who accidentally hacked 7,000 Romo robovacs

    March 6, 2026
    Facebook X (Twitter) Instagram
    Subscribe
    Technology Mag
    Facebook X (Twitter) Instagram YouTube
    • Home
    • News
    • Business
    • Games
    • Gear
    • Reviews
    • Science
    • Security
    • Trending
    • Press Release
    Technology Mag
    Home » A Single Poisoned Document Could Leak ‘Secret’ Data Via ChatGPT
    Security

    A Single Poisoned Document Could Leak ‘Secret’ Data Via ChatGPT

    News RoomBy News RoomAugust 7, 20253 Mins Read
    Facebook Twitter Pinterest LinkedIn Reddit WhatsApp Email
    A Single Poisoned Document Could Leak ‘Secret’ Data Via ChatGPT

    The latest generative AI models are not just stand-alone text-generating chatbots—instead, they can easily be hooked up to your data to give personalized answers to your questions. OpenAI’s ChatGPT can be linked to your Gmail inbox, allowed to inspect your GitHub code, or find appointments in your Microsoft calendar. But these connections have the potential to be abused—and researchers have shown it can take just a single “poisoned” document to do so.

    New findings from security researchers Michael Bargury and Tamir Ishay Sharbat, revealed at the Black Hat hacker conference in Las Vegas today, show how a weakness in OpenAI’s Connectors allowed sensitive information to be extracted from a Google Drive account using an indirect prompt injection attack. In a demonstration of the attack, dubbed AgentFlayer, Bargury shows how it was possible to extract developer secrets, in the form of API keys, that were stored in a demonstration Drive account.

    The vulnerability highlights how connecting AI models to external systems and sharing more data across them increases the potential attack surface for malicious hackers and potentially multiplies the ways where vulnerabilities may be introduced.

    “There is nothing the user needs to do to be compromised, and there is nothing the user needs to do for the data to go out,” Bargury, the CTO at security firm Zenity, tells WIRED. “We’ve shown this is completely zero-click; we just need your email, we share the document with you, and that’s it. So yes, this is very, very bad,” Bargury says.

    OpenAI did not immediately respond to WIRED’s request for comment about the vulnerability in Connectors. The company introduced Connectors for ChatGPT as a beta feature earlier this year, and its website lists at least 17 different services that can be linked up with its accounts. It says the system allows you to “bring your tools and data into ChatGPT” and “search files, pull live data, and reference content right in the chat.”

    Bargury says he reported the findings to OpenAI earlier this year and that the company quickly introduced mitigations to prevent the technique he used to extract data via Connectors. The way the attack works means only a limited amount of data could be extracted at once—full documents could not be removed as part of the attack.

    “While this issue isn’t specific to Google, it illustrates why developing robust protections against prompt injection attacks is important,” says Andy Wen, senior director of security product management at Google Workspace, pointing to the company’s recently enhanced AI security measures.

    Share. Facebook Twitter Pinterest LinkedIn WhatsApp Reddit Email
    Previous ArticleMeta’s prototype headsets show off the future of mixed reality
    Next Article US military finds a good use for Tesla Cybertruck: missile target practice

    Related Posts

    Cloudflare Has Blocked 416 Billion AI Bot Requests Since July 1

    Cloudflare Has Blocked 416 Billion AI Bot Requests Since July 1

    December 6, 2025
    The Louisiana Department of Wildlife and Fisheries Is Detaining People for ICE

    The Louisiana Department of Wildlife and Fisheries Is Detaining People for ICE

    December 5, 2025
    Your Data Might Determine How Much You Pay for Eggs

    Your Data Might Determine How Much You Pay for Eggs

    December 4, 2025
    Russia Wants This Mega Missile to Intimidate the West, but It Keeps Crashing

    Russia Wants This Mega Missile to Intimidate the West, but It Keeps Crashing

    December 4, 2025
    This Hacker Conference Installed a Literal Antivirus Monitoring System

    This Hacker Conference Installed a Literal Antivirus Monitoring System

    December 4, 2025
    Flock Uses Overseas Gig Workers to Build Its Surveillance AI

    Flock Uses Overseas Gig Workers to Build Its Surveillance AI

    December 4, 2025
    Our Picks
    The Corvette ZR1X hybrid can outpace million-dollar sports cars for a fraction of the cost

    The Corvette ZR1X hybrid can outpace million-dollar sports cars for a fraction of the cost

    March 7, 2026
    DJI will pay K to the man who accidentally hacked 7,000 Romo robovacs

    DJI will pay $30K to the man who accidentally hacked 7,000 Romo robovacs

    March 6, 2026
    Grammarly is using our identities without permission

    Grammarly is using our identities without permission

    March 6, 2026
    Valve’s Steam Machine may not launch this year

    Valve’s Steam Machine may not launch this year

    March 6, 2026
    • Facebook
    • Twitter
    • Pinterest
    • Instagram
    • YouTube
    • Vimeo
    Don't Miss
    The Trump administration says it can’t process tariff refunds because of computer problems News

    The Trump administration says it can’t process tariff refunds because of computer problems

    By News RoomMarch 6, 2026

    The US Customs and Border Protection says it currently can’t comply with an order to…

    You can already save up to  on the new M4 iPad Air

    You can already save up to $50 on the new M4 iPad Air

    March 6, 2026
    The Trump phone was a no-show at the world’s biggest mobile show

    The Trump phone was a no-show at the world’s biggest mobile show

    March 6, 2026
    This phone starts fires on purpose

    This phone starts fires on purpose

    March 6, 2026
    Facebook X (Twitter) Instagram Pinterest
    • Privacy Policy
    • Terms of use
    • Advertise
    • Contact
    © 2026 Technology Mag. All Rights Reserved.

    Type above and press Enter to search. Press Esc to cancel.