chiprook
← Security
SecurityOctober 12, 2026, 10:20

Anthropic: AI agents harvested thousands of credentials in six hours

Anthropic's September 2026 threat-intelligence report describes agent-driven attacks that harvested thousands of third-party credentials in about six hours using the victim's own cloud. The GTG-50014 cluster's infrastructure managed more than 23,800 harvested keys and API keys, including AI model credentials.

Anthropic: AI agents harvested thousands of credentials in six hours
#Anthropic#GitHub#Azure#Telegram
Read next
Security

Agent skill harvesting browser credentials had 60,000 GitHub stars

Security

September 2026 Security Review: Attacks via Legitimate Features and AI Execution

Security

Storm-3068 hijacked account via SSPR and stole Kubernetes credentials

Security

Mass-Scanning Campaign Exploits Vite Flaw to Extract Cloud Credentials From Exposed Dev Servers