Lighthouse AdvisorySLED AI Adoption Intelligence
← Back to results

From the NVIDIA edition of October 6, 2026

Vendor claimCautionaryNew this fortnight

NVIDIA bulletin 5903: Model-Optimizer unsafe deserialization (CVE-2026-65142, CVSS 7.8 High), fixed in 0.44.0; first GitHub-only bulletins also include TensorRT 5891

NVIDIA PSIRT · Technology vendor (product security) · Global

NVIDIA
Publisher
GitHub — NVIDIA/product-security
Original publication
2026-10-06
Source retrieved
2026-10-07
Event date
2026-10-06
Read original source

What happened

Bulletin 5903 (Oct 6, 2026) reports CVE-2026-65142 in NVIDIA Model-Optimizer: an attacker may cause deserialization of untrusted data, possibly leading to code execution, data tampering, denial of service and information disclosure. CVSS 3.1 base score 7.8 (High), vector AV:L/AC:L/PR:N/UI:R; affected versions 0 through 0.43.0, fixed in 0.44.0; found externally. The same day, bulletin 5891 reported CVE-2026-65122 in TensorRT, an out-of-bounds read leading to denial of service, CVSS 5.5 (Medium), affecting versions 0 to 11.1, fixed in 11.2. These are the first two bulletins dated after the Oct 1 GitHub-only change.

Why it matters

University research groups and agency AI teams that quantize or optimize models with Model-Optimizer, or build inference engines with TensorRT, often install these from GitHub or pip outside central patch tools. Loading an untrusted model or checkpoint is the likely exposure path for the deserialization flaw (interpretation).

Evidence and measured results

Fetched the Markdown versions of bulletins 5903 and 5891 from the repository and read them in full, including affected and fixed versions.

Limitations and uncertainty

Vendor bulletin; no exploitation details or known-exploited status given. Exposure path is our interpretation.

Cross-source patterns this evidence supports

Put this evidence to work

Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-07; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.

Sales

Role takeaway
Customer problem
AI developer tools in labs and agency pipelines are often unpatched because they sit outside central IT.
Stakeholders
  • Research computing security lead
  • Lab managers
  • Agency AI program lead
Discovery questions
  • Where are Model-Optimizer and TensorRT installed?
  • Do AI environments have SBOMs?
  • Who approves model and checkpoint sources?
Value hypothesis
Inventorying AI SDKs shortens response time for future bulletins (interpretation).
Potential engagement
A one-week AI software inventory and patch sweep across managed images.
Unsupported claims
  • Active exploitation
  • Remote unauthenticated attack (vector is local with user interaction)

Pre-sales engineering

Role takeaway
Fit
Applies wherever NVIDIA Model-Optimizer or TensorRT is used.
Architecture and integration
Scan containers, notebooks and CI images; update to Model-Optimizer 0.44.0 and TensorRT 11.2.
Prerequisites
  • Package inventory or SBOM
  • Image rebuild pipeline
Constraints
  • Pinned research environments
  • Unmanaged researcher machines
Security
Block untrusted checkpoints; patch High-severity deserialization flaw promptly.
Proof of value
  • Scan for affected versions before and after patching

Delivery

Role takeaway
Work
Run a patch campaign for NVIDIA AI SDKs and set model-intake rules.
Dependencies
  • Inventory tooling
  • Lab cooperation
  • Image pipeline
Ownership
Research computing security lead with lab managers.
Skills and adoption
Researchers learn to check NVIDIA bulletins for SDKs they install.
Governance checkpoints
  • Inventory scan
  • Patch completion
  • Image rebuild
Acceptance criteria
  • No affected versions in managed images within 30 days (proposed)
Risks
  • Unmanaged laptops
  • Version pinning
  • Loading untrusted models

Implementation considerations

Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.

Architecture and integration

What must connect, and where does the AI sit in the workflow?

Developer and research environments need software inventories that include Python packages and SDKs, not just drivers.

Governance

Who approves, reviews and stays accountable for outcomes?

Model and checkpoint intake rules (trusted sources only) reduce exposure to deserialization flaws.

Security and privacy

What data, permissions and controls need testing?

High-severity code-execution risk requires user interaction (UI:R) and local access per the vector; update to 0.44.0 and avoid loading untrusted artifacts.

Accessibility and workforce

Who is affected, and what skills or accommodations follow?

Researchers who install their own tools need simple guidance on checking and updating versions.

Procurement

What should contracts, pricing and exit terms secure?

AI platform contracts should state who patches bundled NVIDIA SDKs and how fast.

Operating model

Which teams own the service once it runs?

AI developer environments need to be in scope for vulnerability management.

Publication history

  1. 2026-10-06NVIDIA · Issue 319 resources
Read preserved resource versions (JSON)

Stable resource ID: nvidia-security-bulletin-5903-model-optimizer-cve-2026-65142-20261006