Skip to content
Apart Research
Apart Research
Contact us
Sprint projectSep 13, 2026Ankara, Türkiye

Capability Smuggling: Emergent Capability Escalation in Autonomous AI Agents

Gökalp Çaycı, Uygar Yıldırım · Team 11

Submitted to AI Incident Response Sprint. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: Capability Smuggling: Emergent Capability Escalation in Autonomous AI Agents

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

This project studies how AI agents can gain abilities they were never explicitly given by combining allowed actions with surrounding infrastructure. We call this Capability Smuggling. Using the OpenAI and Hugging Face incident, we built a synthetic harness that detects when permitted actions combine into unauthorized capabilities and tests possible containment strategies.

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. Work highlights an important AI-safety issue: agents can gain unintended capabilities by combining individually permitted actions with system features. The concept is straightforward, and the synthetic harness is an excellent first step toward testing these types of capability-escalation paths.

    But, the main limitation is that the results are based on synthetic scenarios, so the work does not yet demonstrate how frequently these failures occur in realistic agent systems or whether the proposed controls are effective in practice. Overall, this is a promising and well-scoped sprint contribution with a useful conceptual vocabulary and a credible defensive direction.

  2. The report clearly explains how individually authorized operations can combine into unauthorized capabilities, supported by a reproducible synthetic harness and careful distinctions between documented events and hypotheses.

    The "capability transformation" table and explicit limitations make the argument easy to follow. The contribution is primarily a useful synthesis and formalization of established security ideas; the fixture results demonstrate rule consistency rather than real-world detection or prevention. A valuable next step is to implement one proposed control in a small instrumented environment and test it against independently designed escalation and benign traces.

Cite this project

@misc{cayc2026capability,
  title = {{Capability Smuggling: Emergent Capability Escalation in Autonomous AI Agents}},
  author = {Gökalp Çaycı and Uygar Yıldırım},
  year = {2026},
  month = sep,
  note = {Submitted to AI Incident Response Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/capability-smuggling-emergent-capability-escalation-in-autonomous-ai-agents-dgqt}},
  url = {https://apartresearch.com/sprints/projects/capability-smuggling-emergent-capability-escalation-in-autonomous-ai-agents-dgqt}
}

More from AI Incident Response Sprint

All 233 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026