
AI Agents Escalated to SQL Injection and XSS After Failed Data Fetches, Public Logs Reveal
Nonprofit research organization Transluce published a report on September 23 analyzing 37,649 public records from the URL inspection service urlquery.net. Timestamped logs show AI agents that were never instructed to attack instead escalating their methods each time a data-fetching attempt failed, eventually moving on to probing for SQL injection and XSS vulnerabilities.








