Skip to content
Apart Research
Apart Research
Contact us
Sprint projectMay 25, 2026Tokyo, Japan

MicroVMM: AI-Assisted Verification-Oriented Virtual Machine Monitor

Thanh Do · Team BabyPhD

Submitted to The Secure Program Synthesis Hackathon. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: MicroVMM: AI-Assisted Verification-Oriented Virtual Machine Monitor

Open in a new tab (opens in new tab)
Presentation

Presentation: MicroVMM: AI-Assisted Verification-Oriented Virtual Machine Monitor

Open in a new tab (opens in new tab)
Code (opens in new tab)
Share

Modern agentic systems increasingly execute untrusted AI-generated code inside lightweight sandboxing environments. While many existing sandboxes rely on operating-system namespaces, recent advances in autonomous vulnerability discovery have exposed weaknesses in these traditional isolation boundaries. At the same time, vulnerabilities in virtualization infrastructure itself demonstrate that lightweight VMMs remain difficult to reason about formally due to complex shared-memory device protocols such as VirtIO. This project explores whether AI-assisted workflows can help construct formally analyzable virtualization software centered around explicit protocol invariants. Motivated by CVE-2026-5747, we derive queue lifecycle invariants from the vulnerability and construct a minimal verification-oriented VMM in Lean 4. We define abstract operational semantics for queue lifecycle behavior and establish refinement proofs connecting executable implementation behavior to the abstract model. The resulting system proves several safety properties, including post-activation queue immutability, directly eliminating the motivating vulnerability class within the modeled semantics. Our results suggest that AI-assisted development workflows can reduce the engineering complexity of refinement-oriented systems verification when combined with intentionally verification-oriented architectural design.

Reviews

Judging this Sprint?

Review this project

Your public critique appears on this page without your name. Your private critique is not published; only the Apart team reads it. If you agree below, we share your review with grantmaking.ai (opens in new tab) and the Transformative AI Fund so strong projects can be funded.

Not shown on this page.

Shown on this page, without your name.

Only the Apart team reads this, and funders if you agree below.

Share my name publicly on grantmaking.ai *
Share my private critique with funders *

How much would this matter for AI safety if it worked? How innovative is it? For scores of 4-5: is this actually new to the field, or replicating recent work?

Scoring guide
  1. 1Negligible. No clear problem addressed, or no meaningful novelty.
  2. 2Limited. Addresses a real problem but with a generic or well-trodden approach. Incremental at best.
  3. 3Moderate. Clear problem with a reasonable approach; some novelty in framing or method beyond routine application of existing tools.
  4. 4Significant. Important problem with an original approach, or identifies a neglected problem area. A valuable contribution others could build on.
  5. 5Exceptional. Tackles a critical AI safety problem with a genuinely novel approach, or opens a new research direction. Clear theory of change. You'd be excited to share this with researchers in the area.

How sound are methodology, implementation, and findings?

Scoring guide
  1. 1Seriously flawed. Methodology broken, results uninterpretable, or implementation doesn't work.
  2. 2Weak. Approach has significant gaps: missing validation, flawed experimental design, or incomplete implementation.
  3. 3Competent. Technically solid given the short duration. Methodology makes sense, results are interpretable, limitations acknowledged, work builds toward clear conclusions.
  4. 4Strong. Thorough methodology with convincing validation. Results clearly support conclusions. Immediately useful for future work.
  5. 5Exceptional. Ambitious scope executed rigorously. Surprising findings, novel methods, or unusually robust validation.

How clearly are work, findings, and impact potential communicated?

Scoring guide
  1. 1Incomprehensible. Cannot determine what the project is actually claiming or doing.
  2. 2Hard to follow. Key information buried, missing, or diluted by excessive length. Significant effort to extract main points.
  3. 3Clear enough. Can understand the problem, approach, and results without undue effort. Core content clearly present: problem, method, findings, limitations.
  4. 4Well presented. Easy to follow, well-structured, appropriate level of detail. Target audience would get it quickly.
  5. 5Exceptionally clear. A pleasure to read. Complex ideas made accessible. Could serve as a model for how to present this type of work.

  1. While the effort to boot a Linux guest using Lean 4 is great for a hackathon, the paper’s claims go well beyond what is actually demonstrated. Much of the security framing relies on broad AI-safety language, while the core result is a limited systems-verification exercise. Delegating KVM interactions to an unverified C shim leaves a gap in the trusted computing base, so the proofs do not cover the security‑relevant host–guest boundary. Expanding the model to include the FFI boundary, concurrency, and more realistic devices would strengthen the contribution.

  2. It would be helpful to situate this work in related verified OS style work like CertiKOS and Sel4.

    The claim is that AI-Assistance can help to develop verified vmms, but all we are given is an AI generated github repo. While this is in a way an existence proof, there is no baseline (how long does it take to generate this without AI). There is also no clear demonstration that the CVE of interest is no longer viable on this system. Additionally, how should we trust that the properties proved in Lean are actually useful? The generated documentation is so verbose it is not clear what we should care about.

Cite this project

@misc{do2026microvmm,
  title = {{MicroVMM: AI-Assisted Verification-Oriented Virtual Machine Monitor}},
  author = {Thanh Do},
  year = {2026},
  month = may,
  note = {Submitted to The Secure Program Synthesis Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/microvmm-aiassisted-verificationoriented-virtual-machine-monitor-qcab}},
  url = {https://apartresearch.com/sprints/projects/microvmm-aiassisted-verificationoriented-virtual-machine-monitor-qcab}
}

More from The Secure Program Synthesis Hackathon

All 61 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026