Amazon says attackers' AI agents exploited MONAI Label zero-day in about 11 minutes
For the ciso of a regional bank
Amazon said Friday that attackers' AI agents independently found a zero-day in MONAI Label, an open-source medical imaging AI application, and used it to write files outside its storage directory in about 11 minutes.1
Amazon found the attackers' Cairn server on Sept. 10 running without authentication and read the operator's prompts, which gave the agents one goal: "obtain shell access."1 Cairn is an open-source framework that coordinates AI agents for security testing, and Amazon assesses with high confidence that the flaw was a zero-day because it found no prior fix, issue, advisory or mention in vulnerability databases.1
According to Amazon, the campaign ran about 48 hours from Sept. 8 against universities in East Asia outside mainland China, and Amazon assesses a China nexus with moderate confidence.12
The agents abused a file-upload feature to place files outside the intended storage area, then in a later session ran malicious code and gained full control of the server; The Register reported, citing Amazon CISO CJ Moses, that this was a Python reverse shell running as root.12
The fix was merged into MONAI Label's main branch on Oct. 19 in pull request #1915, but the newest release on PyPI is still 0.8.5, uploaded Nov. 25, 2024, and the project's advisory lists no patched version for 0.8.5 and earlier.1
Authentication is disabled by default, according to the advisory, so anyone who can reach the server can exploit the flaw, and users with the ANNOTATOR role can exploit it even with authentication enabled.1
Banks and fintechs running MONAI Label, for example in health-related lending, insurance or research partnerships, cannot patch by upgrading and must rely on network isolation and access controls until a release ships. The reported targets were universities, not banks, but the same agent-driven approach to finding flaws is now documented in the wild, so teams should check for earlier compromise and not assume turning on authentication fixes it.
Administrators should watch for a MONAI Label release containing the fix, which the project's advisory says will be in the next release; the sources give no date.1
Our take. Attacker agents needed about 11 minutes to find and use the flaw. The merged fix has waited for a release since Oct. 19. For open-source tools you expose, the patch is the slow part, so cut exposure first.
Action. Vulnerability management: Inventory internet-facing open-source servers with authentication off by default, and restrict network access now instead of waiting for releases.
Sources
- ReportAmazon says attackers' AI agents exploited a zero-day in 11 minutes Venturebeat
- Also reported byChina attackers deployed AI agents that autonomously found and exploited a 0-day to take over a university server Theregister