Server-management flaws expose AI workflow platforms and data centers

Researchers have disclosed six remote‑code‑execution (RCE) vulnerabilities in Flowise, an open‑source platform used to build AI agents and automated workflows. The flaws affect Flowise versions 3.1.1 and 3.1.2 and can be exploited by authenticated users through components such as the CSVAgent, the JavaScript sandbox (vm2), and custom MCP configurations. The weaknesses could allow attackers to run arbitrary commands on the host server, potentially compromising data, credentials and any connected business tools.

A separate study by the security firm Lava scanned the public internet for exposed Baseboard Management Controllers (BMCs) that use the IPMI 2.0 protocol. The scan identified roughly 37,000 reachable BMCs, with about two‑thirds vulnerable to the long‑standing CVE‑2013‑4786 authentication flaw. Captured password hashes were often easily cracked, giving attackers the ability to power‑cycle servers, flash firmware and maintain persistence below the operating‑system level. The risk is amplified in AI and GPU cloud environments where many high‑value systems share out‑of‑band management networks.