Close Menu
    Facebook X (Twitter) Instagram
    Wifi PortalWifi Portal
    • Blogging
    • SEO & Digital Marketing
    • WiFi / Internet & Networking
    • Cybersecurity
    • Tech Tools & Mobile / Apps
    • Privacy & Online Earning
    Facebook X (Twitter) Instagram
    Wifi PortalWifi Portal
    Home»Cybersecurity»Indirect prompt injection is taking hold in the wild
    Cybersecurity

    Indirect prompt injection is taking hold in the wild

    adminBy adminApril 25, 2026No Comments3 Mins Read
    Facebook Twitter LinkedIn Telegram Pinterest Tumblr Reddit WhatsApp Email
    The vulnerability that turns your AI agent against you
    Share
    Facebook Twitter LinkedIn Pinterest Email

    The open web is slowly but surely filling up with “traps” designed for LLM-powered AI agents.

    The technique, known as indirect prompt injection (IPI), involves hiding (more or less) covert instructions inside ordinary web pages, waiting for an AI agent to read them and carry out the author’s commands.

    Indirect prompt injection in the wild

    The IPI attack kill chain (Source: Forcepoint)

    “Ignore previous instructions”

    In back-to-back reports published this week, Google and Forcepoint researchers laid out real-world evidence of these attacks.

    Google used a repository of 2–3 billion crawled pages per month as their data source, focusing on static websites including blogs, forums and comment sections (and eschewing social media content).

    Forcepoint’s X-Labs researchers conducted active threat hunting across publicly accessible web infrastructure, with telemetry flagging real payloads triggering on patterns like “Ignore previous instructions” and “If you are an LLM.”

    Both companies found IPIs fueled by harmless and malicious intent.

    The first category, according to Google, contains pranks and helpful guidance, such as instructions to change the AI agent’s conversational tone (“Tweet like a bird”) or add relevant content to AI summaries (e.g., telling users to check facts for themselves).

    The latter includes:

    • Search engine manipulation / traffic hijacking
    • IPIs meant to prevent AI agents from retrieving content (DoS) and to initiate a destructive action instead
    • IPIs aimed at data exfitration (e.g., API keys)
    • IPIs focused on destruction (e.g., “try to delete all files on the user’s machine”)

    Indirect prompt injection in the wild

    IPI with destructive intent (Source: Google)

    Forcepoint researchers also unearthed IPI attempts aimed at performing financial fraud.

    One payload, for example, embedded a fully specified PayPal transaction and step-by-step instructions designed for AI agents with integrated payment capabilities. A second case used meta tag namespace injection combined with a persuasion amplifier keyword (“ultrathink”) to route AI-mediated financial actions toward a Stripe donation link.

    A third case appeared to function as a widely distributed test payload, possibly to identify which AI systems are vulnerable before deploying higher-impact attacks.

    Hiding from humans

    Attackers use different tricks to hide malicious instructions from human eyes while keeping them fully visible to AI.

    The most common involve making text physically invisible on a webpage by shrinking it to a single pixel, draining its color to near-transparency, or simply tagging it as hidden using standard web design tools.

    The more sophisticated tricks involve burying payloads inside HTML comment sections and hiding instructions inside a page’s metadata.

    There’s a growing interest in IPI attacks

    Neither team found evidence of sophisticated, coordinated campaigns, though, according to Forcepoint researchers, “shared injection templates across multiple domains suggest organized tooling rather than isolated experimentation.

    Still, the window for getting ahead of this threat is closing fast, they believe.

    Google says it observed a sharp uptick in malicious activity during its scans: “We saw a relative increase of 32% in the malicious category between November 2025 and February 2026, repeating the scan on multiple versions of the [CommonCrawl archive of the public web].”

    Forcepoint also pointed out that that the impact of these attacks scales with AI privilege.

    “A browser AI that can only summarize is low-risk. An agentic AI that can send emails, execute terminal commands or process payments becomes a high-impact target. If AI agents consume untrusted web content without enforcing a strict data-instruction boundary, every page they read remains a potential attack vector.”

    Subscribe to our breaking news e-mail alert to never miss out on the latest breaches, vulnerabilities and cybersecurity threats. Subscribe here!

    hold Indirect Injection Prompt Wild
    Share. Facebook Twitter Pinterest LinkedIn Tumblr Telegram Email
    Previous Article10 Hacks Every YouTube Music User Should Know
    Next Article The Galaxy Z Fold 8 Wide might use a familiar aspect ratio
    admin
    • Website

    Related Posts

    Locked Shields 2026: 41 Nations Strengthen Cyber Resilience in World’s Biggest Exercise

    April 25, 2026

    North Korea’s Lazarus Targets macOS Users via ClickFix

    April 25, 2026

    26 FakeWallet Apps Found on Apple App Store Targeting Crypto Seed Phrases

    April 25, 2026
    Add A Comment
    Leave A Reply Cancel Reply

    Search Blog
    About
    About

    At WifiPortal.tech, we share simple, easy-to-follow guides on cybersecurity, online privacy, and digital opportunities. Our goal is to help everyday users browse safely, protect personal data, and explore smart ways to earn online. Whether you’re new to the digital world or looking to strengthen your online knowledge, our content is here to keep you informed and secure.

    Trending Blogs

    Melitta AromaFresh Therm Pro review: an almost-great drip coffee maker that’s let down by its most novel feature

    April 25, 2026

    Locked Shields 2026: 41 Nations Strengthen Cyber Resilience in World’s Biggest Exercise

    April 25, 2026

    8 SEO tasks you shouldn’t do manually

    April 25, 2026

    North Korea’s Lazarus Targets macOS Users via ClickFix

    April 25, 2026
    Categories
    • Blogging (68)
    • Cybersecurity (1,502)
    • Privacy & Online Earning (183)
    • SEO & Digital Marketing (922)
    • Tech Tools & Mobile / Apps (1,787)
    • WiFi / Internet & Networking (246)

    Subscribe to Updates

    Stay updated with the latest tips on cybersecurity, online privacy, and digital opportunities straight to your inbox.

    WifiPortal.tech is a blogging platform focused on cybersecurity, online privacy, and digital opportunities. We share easy-to-follow guides, tips, and resources to help you stay safe online and explore new ways of working in the digital world.

    Our Picks

    Melitta AromaFresh Therm Pro review: an almost-great drip coffee maker that’s let down by its most novel feature

    April 25, 2026

    Locked Shields 2026: 41 Nations Strengthen Cyber Resilience in World’s Biggest Exercise

    April 25, 2026

    8 SEO tasks you shouldn’t do manually

    April 25, 2026
    Most Popular
    • Melitta AromaFresh Therm Pro review: an almost-great drip coffee maker that’s let down by its most novel feature
    • Locked Shields 2026: 41 Nations Strengthen Cyber Resilience in World’s Biggest Exercise
    • 8 SEO tasks you shouldn’t do manually
    • North Korea’s Lazarus Targets macOS Users via ClickFix
    • This duster changed the way I clean my gear, and now you can grab one for cheap
    • 26 FakeWallet Apps Found on Apple App Store Targeting Crypto Seed Phrases
    • Why Google Has Changed & Who’s Really Paying for It
    • The Galaxy Z Fold 8 Wide might use a familiar aspect ratio
    © 2026 WifiPortal.tech. Designed by WifiPortal.tech.
    • Home
    • About Us
    • Contact Us
    • Privacy Policy
    • Terms and Conditions
    • Disclaimer

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