Skip to content
Apart Research
Apart Research
Contact us
Sprint projectSep 13, 2026Lagos, Nigeria

Detection Was Not the Failure: A Replayable Conformance Harness for the July 2026 Agent Intrusion1

Chijioke Ubajaka

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

Read the report

Report: Detection Was Not the Failure: A Replayable Conformance Harness for the July 2026 Agent Intrusion1

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

Presentation: Detection Was Not the Failure: A Replayable Conformance Harness for the July 2026 Agent Intrusion1

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

Hugging Face's forensic reconstruction of the July 2026 agent intrusion records that its security stack correlated the attack into a coherent signal and then failed to raise criticality and page the on-call team. I treat that as the primary failure and test it. From published aggregates: per-phase action counts, daily totals and seventeen timestamped anchors, I reconstruct a defender-visible event stream by iterative proportional fitting, generate 30 days of synthetic benign platform workload, and score four escalation policies on detection point and false-page cost. A per-event severity policy never fires. A boundary-accumulation policy pages 51.5 hours before containment but costs 69 false pages per week, driven by the deploy pipeline. A principal-scoped variant retains that lead at 0.5 false pages per week, but only while the principal-to-boundary map stays accurate to better than one violation in ten thousand sessions. The bottleneck is identity hygiene, not detection logic.

Reviews

Judging this Sprint?

Review “Detection Was Not the Failure: A Replayable Conformance Harness for the July 2026 Agent Intrusion1”

Your public critique appears on the project page without your name. Reviews of strong projects are shared with funders, as you choose below.

Not shown on the project page.

Shown on the project page, without your name.

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

Decline to score

Only when the quality of the writing makes the work impossible to follow. Say why in your private critique; the project page shows only that a judge declined, not why.

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.
Sharing your name *

When we share a strong project with funders, your name, scores and public critique go with it, confidentially; this choice is only about whether your name is public. If you think this project deserves funding, please also endorse it: that carries the most weight.

Share my private critique with funders *

See our Privacy Policy.

  1. This is a strong submission, and unusually reproducible. Turning a real incident into a testable detection experiment is genuinely hard to pull off, and the comparison between detection lead time and the cost of false pages is the most valuable part. The unreconciled event counts are handled up front instead of glossed over, and the pinned timestamps, the mix of benign and attack traffic, the sensitivity testing, regression guards, and the honest limitations section all make the results easier to trust.

    The main gap is external validity. The benign workload, defender visibility, session linkage, and map drift rate are all synthetic or assumed, and the principal scoped policy only works if the identity map stays extremely accurate, which is a big ask in practice. A good next step would be validating against even a small sample of real operational telemetry, and spelling out how event severity and rate baselines were assigned. It would also be worth testing identity rotation, fragmented sessions, and activity timed around legitimate deployment windows, to see whether the approach holds up against an adaptive agent.

    Read full reviewShow less

  2. This study gives containment discussions a useful testbed by comparing warning time with the cost of false alerts. Including legitimate deployments that cross trust boundaries makes the comparison more realistic, and the report explains its assumptions clearly. The findings are specific to the simulated workload, with the severity choices and hourly alert timing limiting the policy comparison. The released harness makes those assumptions inspectable and gives the work value beyond its headline numbers.

Cite this project

@misc{ubajaka2026detection,
  title = {{Detection Was Not the Failure: A Replayable Conformance Harness for the July 2026 Agent Intrusion1}},
  author = {Chijioke Ubajaka},
  year = {2026},
  month = sep,
  note = {Submitted to AI Incident Response Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/detection-was-not-the-failure-a-replayable-conformance-harness-for-the-july-2026-agent-intrusion1-pp2o}},
  url = {https://apartresearch.com/sprints/projects/detection-was-not-the-failure-a-replayable-conformance-harness-for-the-july-2026-agent-intrusion1-pp2o}
}

More from AI Incident Response Sprint

All 231 entriesBrowse all projects

Build something like this at the next Sprint

AI x Epistemics Research Sprint · Nov 13 - 15, 2026