New AI-Powered Threat Allows Hackers to Gain AWS Admin Access in Minutes

A sophisticated, AI-accelerated cloud intrusion targeted an AWS environment: threat actors discovered exposed S3 credentials (with AI-related naming), used LLMs to generate and inject malicious Python into a Lambda to create admin keys, escalated to administrative privileges in under 10 minutes, moved laterally across 19 principals, deployed a Terraform-backed Lambda for Bedrock credential theft, launched a p4d.24xlarge GPU instance and installed a JupyterLab backdoor to persist and steal compute; the report concludes with mitigation guidance (eliminate long-term credentials, secure S3, restrict Lambda permissions, and monitor Bedrock usage).