Rising Threat: Malicious AI Agents in Open Source Repositories
PremiumESET finds an increase in malicious AI agents in open-source repos designed to trick developers and steal secrets.
§Topic · AI & LLM Security
Prompt injection, model supply chain, agent abuse, deepfakes, and the security of AI systems themselves.
All dispatchesESET finds an increase in malicious AI agents in open-source repos designed to trick developers and steal secrets.
AI Now Institute's 'Friendly Fire' shows autonomous AI coding agents can be tricked into executing attacker code during self-approved scans.
Wiz-discovered GhostApproval symlink flaw lets six AI coding assistants run attacker-controlled code by tricking file approvals.
AI gateway compromise led to cryptomining, demonstrating exposure of models, cloud resources, and IAM data via gateway attacks.
Sygnia analysis shows agentic AI compressed cloud intrusion timelines to 72 hours, emphasizing speed and orchestration risk.
WriteOut session isolation bug in Writer AI could leak session tokens across tenants enabling full account takeover.
Researchers report JadePuffer, an agentic ransomware automating attack steps and accelerating operations.
Get these articles delivered to your inbox.
Subscribe free