Indirect prompt injection found in webpage HTTP headers
Signals Corps sampled 217 million HTTP responses from Common Crawl and found nearly 8,000 sites messaging AI agents via headers. Thirty sites attempted indirect prompt injection; most were pranks, but some could have influenced visiting agents.
- 217 million HTTP responses from Common Crawl analyzed
- About 8,000 sites message AI agents via headers
- 30 sites attempted indirect prompt injection
- 6,989 sites sent the x-llms-txt header
Read next
Security