NVIDIA Patches High-Severity “Insecure Deserialization” Flaws in BioNeMo Framework

NVIDIA has issued an important security update for its BioNeMo Framework, a critical tool used by researchers and developers in the generative AI and drug discovery space. The update addresses two high-severity vulnerabilities that could allow an attacker to seize control of systems or disrupt vital research operations.

The core of the threat lies in how the framework handles incoming data. According to the security bulletin, NVIDIA BioNeMo contains a vulnerability where a user “could cause a deserialization of untrusted data”.

In the world of cybersecurity, “deserialization” is the process of taking data from a file or network and turning it back into a functional object in memory. If a program doesn’t carefully check that data first, an attacker can “poison” it with malicious code.

As NVIDIA warns, a successful exploit of this flaw “might lead to code execution, denial of service, information disclosure, and data tampering.”

Two specific vulnerabilities are being addressed in this software update:

• CVE-2026-24164: This is the more dangerous of the two, carrying a CVSS score of 8.8 (High). It can be exploited over a network, meaning an attacker doesn’t need physical or local access to the target machine.
• CVE-2026-24165: With a CVSS score of 7.8 (High), this vulnerability also centers on the deserialization of untrusted data. While it carries a slightly lower score, its potential impact on data integrity and system availability remains a serious concern.

Both flaws are categorized under CWE-502, a well-known weakness involving the processing of untrusted data during deserialization.

The vulnerability affects all Linux-based versions of the BioNeMo Framework that do not include a specific security fix. NVIDIA is urging all users to take immediate action.

To protect your system, clone or update this software to include commit e5e58c8 or later from the NVIDIA/BioNeMo Framework GitHub repo.