Unknown attackers compromised LiteLLM's release process on Wednesday, distributing a malicious version (1.107.0) through PyPI and Docker registries containing malware designed to steal cloud and AI credentials plus a remote access trojan capable of executing arbitrary code. Users who ran the compromised version should treat their secrets as compromised.
Security firm PromptArmor has demonstrated that Atlassian's Rovo AI assistant can be hijacked through hidden instructions in uploaded PDFs or malicious links to silently exfiltrate sensitive data from Jira, Confluence, and third-party integrations to attacker-controlled servers without user approval or audit trails.