OpenHands fixes critical 9.8-severity command-execution flaw
Versions below 1.52.0 let unauthenticated attackers run administrator-level commands in the AI agent's workspace, dailysecu.com reported.
OpenHands, an AI software-development agent tool, has patched a critical vulnerability tracked as CVE-2026-112346, according to dailysecu.com. The flaw carries a CVSS 3.1 score of 9.8, and the fix ships in version 1.52.0.
All versions of the OpenHands agent development tool below 1.52.0 are affected. The vulnerability allows an unauthenticated attacker to run arbitrary commands with administrator privileges inside the agent's working environment.
The exposure stems from how OpenHands handles its work environments: when running Docker and Apptainer environments, the tool binds its AI agent server to all network interfaces, and authentication could be disabled by default.
Workspace data at risk
Successful exploitation could expose source code, files, credentials and secrets stored in the development workspace, according to the published information.
The published information contains no evidence that the vulnerability has been exploited in real attacks. Users are directed to version 1.52.0, which resolves the issue.
Quick answers
What is CVE-2026-112346?
It is a critical command-execution vulnerability in the OpenHands AI software-development agent tool, with a CVSS 3.1 score of 9.8.
Which OpenHands versions are affected?
Versions below 1.52.0 are affected; version 1.52.0 contains the fix.
Has the flaw been exploited in attacks?
The published information contains no evidence of exploitation in real attacks.