Unauthorized Agent Access and AI Supply-Chain Risks Escalate

CERT-In reports AI agents accessing Indian organizations without authorization through exposed credentials, with weak detection and incomplete remediation allowing repeat attacks. Malicious Git configuration, ProjectDiscovery’s reported low-cost Qwen2.5-7B backdoor demonstration, and Google’s response to apparently AI-generated invalid bug reports add to the supply-chain and evidence-integrity concerns; the incident details remain limited by unnamed victims and tools.