Attackers Used AI to Breach an AWS Environment in 8 Minutes

Sysdig researchers describe a late-2025 cloud compromise where attackers abused credentials found in misconfigured public S3 buckets to perform Lambda code injection, escalate to administrative privileges in roughly eight minutes, exfiltrate secrets and configuration data, and attempt GPU hijacking and abuse of cloud LLM services; the operation showed indicators of LLM-assisted automation and rapid, iterative exploitation, prompting recommendations for least-privilege controls and runtime detection.