Hackers Use Hidden Website Instructions in New Attacks on AI Assistants
Forcepoint discovers indirect prompt injection attacks using hidden website code to exploit AI assistants.
Cybersecurity researchers at Forcepoint X-Labs have confirmed that Indirect Prompt Injection (IPI) attacks are now active in the wild, with hackers hiding malicious commands in website code using techniques like invisible fonts, HTML comments, and CSS tricks to exploit AI assistants including GitHub Copilot and Claude Code. The research identified ten real-world examples of IPI attacks deployed across websites in April 2026, targeting use cases including financial fraud, API key theft, data deletion, and DoS attacks. Attackers exploit the AI's inability to distinguish between data and instructions, allowing them to manipulate autonomous agents without direct user involvement.
AI summary of the linked source. Not reviewed by an editor. How this works
Run Forcepoint?
Get an email when a new story names Forcepoint, usually within the hour.
Free. Your list stays private and never appears in a subject line. One click stops it. How Vendor Watch worksPrivacy
What helps against this
How we list vendors
- DNS filtering and secure web gateway
Users cannot reach known-malicious sites, and malware cannot phone home.
- AI application security
Prompt injection, data leaks through AI tools and unapproved AI use get caught.
- Security awareness training and phishing simulation
People recognise and report attacks instead of falling for them.
Indicators of Compromise
Extracted by AI from the linked source. Not reviewed by an editor. How this works
- domain — faladobairro.com
- domain — perceptivepumpkin.com
- domain — thelibrary-welcome.uk
- domain — bentasker.co.uk
- domain — lcpdfr.com
- domain — kleintechnik.net
- domain — kassoon.com