Skip to content
Apart Research
Apart Research
Contact us
Sprint projectNov 23, 2025Durham, UK

Ghost Marks in the Machine: A Critical Review of SynthID for Code Provenance Monitoring

Eve Sherratt-Cross, Theo Farrell, Sam Ogden, Oscar Ryley · Team Durham AI Safety Initiative

Submitted to Defensive Acceleration Hackathon. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: Ghost Marks in the Machine: A Critical Review of SynthID for Code Provenance Monitoring

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

AI-generated code is increasingly common in software and prone to security vulnerabilities. It is hence critical to monitor the origins of code used in secure applications. SynthID is a Google DeepMind method for watermarking AI-generated text, images and videos but there is currently no existing mechanism for code. We adapt various SynthID schemes to Python code, and analyse how effective they are using Bayesian detectors. We find that longer n-grams support more robust watermark detection, but the corresponding generated code is more prone to syntax and runtime errors. This work paves the way for critical future work, because code origin monitoring forms part of robust cyber defences against vulnerable or backdoored AI-generated code.

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 *

Does this reduce AI-related catastrophic or existential risks?

Scoring guide
  1. 1Minimal Impact. The project has minimal relevance to AI safety. It doesn't meaningfully address risks uniquely enabled or accelerated by advanced AI systems.
  2. 2Tangential Connection. The project touches on AI safety concepts but lacks depth or specificity. The connection to AI-enabled threats (bio, cyber, or AI misuse) is weak or unclear.
  3. 3Clear AI Safety Value. The project clearly reduces AI-related risks with valuable contributions. It addresses specific threats from AI systems and engages meaningfully with biosecurity, cybersecurity, or AI safety challenges.
  4. 4Significant Impact Potential. The above, plus the project demonstrates scalable safety mechanisms or defensive approaches. It shows clear potential to buy time for solving harder problems like alignment, or creates positive externalities for the broader AI safety ecosystem.
  5. 5Major Advancement. The above, plus the project represents a significant leap forward in defensive AI safety. Judges would eagerly share this with biosecurity, cybersecurity, or AI safety researchers and expect it to influence the field.

Does this strengthen the shield against AI-enabled threats?

Scoring guide
  1. 1Minimal Relevance. The project is only tangentially related to defensive technology or societal protection. Connection to biosecurity, cybersecurity, or defensive infrastructure is unclear or missing.
  2. 2Some Relevance. The project has some relevance to defensive acceleration, but the connection is broad or generic. It touches on defense without specific focus on AI-enabled threats or protective capabilities.
  3. 3Clear Relevance. The project clearly addresses defensive gaps against AI-enabled threats. It connects to at least one track (biosecurity, cybersecurity, or defense infrastructure) and demonstrates understanding of the threat landscape.
  4. 4Strong Contribution. The above, plus the project builds on existing defensive approaches and offers novel tools, frameworks, or implementations. It explicitly explains how it strengthens defensive capabilities with realistic deployment potential.
  5. 5Breakthrough Impact. The above, plus the project provides breakthrough insights or tools that could significantly influence defensive technology development. It identifies critical gaps and presents compelling solutions with clear paths from prototype to deployed system.

Did you build something that actually works?

Scoring guide
  1. 1Incomplete or Flawed. The project appears rushed or incomplete. Technical implementation is flawed, core functionality doesn't work, or the approach is fundamentally unsound. Little to no documentation.
  2. 2Basic Competence. The project shows reasonable effort with basic technical competence. Core functionality partially works. Documentation exists but may be incomplete. Some limitations are acknowledged.
  3. 3Solid Hackathon Project. The project is technically solid and well-scoped for 48 hours. Core functionality works and is documented. Code/methods are understandable and limitations are honestly addressed. This is what a good weekend prototype should look like.
  4. 4Impressive Implementation. The above, plus the implementation exceeds typical hackathon quality. Clear methodology, thorough documentation, and working demo. The tool/prototype is immediately useful for defenders and could realistically be built upon.
  5. 5Exceptional Execution. The project far exceeds expectations with exceptional technical execution. The implementation is elegant, fully functional, and includes something special (e.g., deployed demo, exceptional documentation, innovative architecture, or clear startup potential).

  1. Strengths: Solid extension research. SynthID doesn't exist for code, and you actually built it out. Execution quality is high for hackathon scope.

    Suggestions: The threat model assumes the bottleneck is identifying which code is AI-generated. But if the underlying concern is vulnerability rates, why does origin matter? The defensive value here seems to be compliance and governance (are developers using unauthorized tools?) rather than security (is this code safe?). Those are different problems. For d/acc framing, we'd want to see why origin-aware review beats origin-agnostic vulnerability scanning and remediation.

    From a Halcyon Ventures investor's POV: There's probably enterprise value in the compliance use case. "Did our contractors use unauthorized AI tools?" is a real question procurement and legal teams ask. But that's governance value, not security value. Worth being precise about which problem you're solving. Great technical hustle, though, and still really cool that you are thinking about how to identify synthetic code in the wake of the SynthID launch!

    Read full reviewShow less

  2. Nice project; it’d be more compelling if you tied it more to concrete security/safety threats and use cases.

  3. Great project name and really solid attempt at addressing an important problem. Adapting SynthID-style watermarking to code is a clever starting point, but the current method still feels too fragile to handle complex or production systems.

    I see value in contexts where an entire organization wants to track which parts of their codebase are AI-generated versus human-written. For example, a company encouraging developers to use AI coding tools might want clear provenance for maintainability and accountability.

    As a broader defense against malicious or low-quality AI code entering the global software ecosystem, this approach seems too easy to evade. It would likely only catch developers who are not intentionally hiding the use of AI and it seems would pretty easily be able to remove the watermarks.

Cite this project

@misc{sherrattcross2025ghost,
  title = {{Ghost Marks in the Machine: A Critical Review of SynthID for Code Provenance Monitoring}},
  author = {Eve Sherratt-Cross and Theo Farrell and Sam Ogden and Oscar Ryley},
  year = {2025},
  month = nov,
  note = {Submitted to Defensive Acceleration Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/ghost-marks-in-the-machine-a-critical-review-of-synthid-for-code-provenance-monitoring-ov2c}},
  url = {https://apartresearch.com/sprints/projects/ghost-marks-in-the-machine-a-critical-review-of-synthid-for-code-provenance-monitoring-ov2c}
}

More from Defensive Acceleration Hackathon

All 84 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026