Loki Intelligence — Security Briefs · Published
Daily Security Brief: ConnectWise Supply Chain, LiteLLM Package Compromise, OpenCode MCP Flaw
Today's reports center on supply chain and AI tooling risk: a ConnectWise Cloud Backup compromise, a malicious LiteLLM release, and an OpenCode flaw tied to MCP servers. Teams should prioritize auditing release pipelines and credentials, patching developer tooling, and assigning clear ownership to AI agents and their access.
Signal 01 · The Hacker News
China-Linked Hackers Use ARTEX AI Agents in ConnectWise Supply Chain Attack
Microsoft disclosed that an actor it tracks as Storm-0940, described as China state-aligned, was behind the recent supply chain compromise of the ConnectWise Cloud Backup application. The actor reportedly used ARTEX, a Go-based agentic AI penetration testing framework, during the incident.
Why it matters: Vendor-delivered software is a trust path into your environment, so monitor third-party backup and management tools for unexpected behavior and review the access they hold.
Source: The Hacker News
Signal 02 · CISA
Siemens PSS(R)SINCAL
CISA published an advisory on Siemens PSS(R)SINCAL, which relies on the third-party WibuKey product that contains an out-of-bounds write vulnerability. Exploitation could crash the system and cause denial of service, and it applies only to certain configurations.
Why it matters: Check whether your deployments use the affected configuration and follow the vendor and CISA guidance for third-party component updates and mitigations.
Source: CISA
Signal 03 · The Hacker News
The AI Agent Nobody Owns: Watch Expert Webinar On Your Newest Insider Threat
A webinar promotion highlights the risk of unmanaged AI agents running on employee machines that hold API keys and database passwords and can reach internal systems. It frames these agents as an insider-threat problem because no one reviewed their access or owns them.
Why it matters: Inventory the AI agents in use, assign an owner to each, and limit the credentials and systems they can reach.
Source: The Hacker News
Signal 04 · Cloudflare Security
Cloudflare deploys protections against php2shell vulnerability
Cloudflare said it has enabled default WAF protections for newly disclosed PHP vulnerabilities, including one referred to as php2shell, ahead of public disclosure. The rules use request and response signals to block exploit attempts and keep exposed data from reaching attackers.
Why it matters: A WAF rule is a stopgap, so PHP-based applications should still be patched and their exposure reviewed.
Source: Cloudflare Security
Signal 05 · The Hacker News
OpenCode Flaw Lets Malicious MCP Servers Execute Code During Sign-In on Windows and WSL
A flaw in the OpenCode AI coding agent let a malicious or hijacked MCP server run code on a developer's Windows or WSL machine during sign-in. OpenCode connects to MCP servers to extend its capabilities.
Why it matters: Update the tool, connect coding agents only to MCP servers you trust and have vetted, and treat MCP connections as part of your attack surface.
Source: The Hacker News
Signal 06 · The Hacker News
LiteLLM Suspects GitHub Token Leak in AI Gateway Logs Behind Malicious Package Release
LiteLLM disclosed that attackers defeated multiple security controls and used its release pipeline on October 21, 2026 to publish a malicious package version containing a credential stealer and a backdoor. The project suspects a GitHub token leaked through AI gateway logs.
Why it matters: Check which LiteLLM version you installed, rotate exposed secrets, and keep tokens out of logs while limiting the scope of release-pipeline credentials.
Source: The Hacker News
Signal 07 · GitHub Security
How bug bounty sharpens one researcher’s security skills
As part of Cybersecurity Awareness Month, the GitHub Bug Bounty team profiled a researcher, covering their research approach, favorite bug classes, and advice for newcomers.
Why it matters: Researcher-driven disclosure channels such as bug bounty programs can supplement internal testing, and teams without one can consider authorized, scoped security testing of their web apps, APIs and AI agents.
Source: GitHub Security
Brief sources
Related briefs
- What Is AI Agent Security Testing?
- AI Red Teaming vs Traditional Pentesting
- MCP Security Testing Checklist
Relevant Loki service: Orvyn — AI agent security private preview.