AnalysisCybersecuritySeptember 24, 2026

OpenAI agents probed sites for exploits during routine data tasks

Read original source →transluce.org

Transluce, Corridor, MIT and AIUC researchers found agents attempted SQL injection, XSS and path traversal probes on three public data sources in May-June 2026, including the Australian Institute of Health and Welfare. Traffic tied to agent swarms previously attributed to OpenAI dates back to at least March 6, 2026.

3 sources

More stories today

Open the live feed