P024 Data in and out of sealed verification environments
Can data be moved into a sealed verification facility, and results out, while bounding what leaks, including through faults and retries?
Protocols for moving data into, deleting it from, and bounding leakage out of a sealed auditing facility (including one-bit outputs and fault budgets).
Terms: SCIF (sensitive compartmented information facility): a shielded, access-controlled room built to stop information leaking out. air-gap: physically isolating a system from outside networks.
Status
Analyzed Design only: a neutrally located, mutually monitored and possibly air-gapped data centre runs privacy-preserving evaluations, assessors emit one bit per judgment, and the sensitive data is verifiably deleted afterwards (Fig. 2.1 caption); how the bit is transferred is left to future work (fn 261). The fault-and-retry leakage budget appears only as a stated requirement in Cankaya's self-deprecated System Overview v0.2 (a pre-agreed budgeted fault rate). Nothing built, no red-team, no measured leak bound.
After conducting their evaluations as agreed by the Prover and Verifier, the assessor would be able to reliably transmit one bit per judgment to the Verifier and Prover.
Raised in: Harack et al. (2025), Verification for International AI Governance (Oxford AIGI); Cankaya (2026), A System Overview for Near-Term, Low-Trust AI Compute Verification (v0.2); Cankaya (2026), All Hands on Deck to Build the Datacenter Lie Detector; Cankaya (2026), Research Note: The Fundamentals and Feasibility of Secure Network Taps for Verifying AI Datacenter Use.
Sub-questions
| ID | Sub-question | What the sources report |
|---|---|---|
| P024.1 | Has a protocol for moving confidential inputs into an air-gapped verification facility (single transmission before split) been red-teamed with zero exfiltration observed? | open |
| P024.2 | Has a protocol been demonstrated by which two parties confirm deletion of data from an attested cluster (e.g. attested key destruction)? | implied by a source, not stated as open |
| P024.3 | Has a two-party 'either can zeroise' control for a server cluster been built, with an attested zeroisation time? | implied by a source, not stated as open |
| P024.4 | Has a one-bit output channel from a sealed facility been specified with a bound on extra bits leaked (e.g. via timing)? | open |
| P024.5 | What fault/retry rate can be pre-agreed so that deliberate faults (each leaking one bit) stay within an acceptable leakage budget, and what is the benign fault rate of a redundant auditing environment? | partly answered |
Sub-questions not yet stated precisely enough to check
| ID | Sub-question |
|---|---|
| P024.6 | How is power supplied to the auditing environment without opening an attack or leak path? |
| P024.7 | Can human assessors be used in sealed facilities given they cannot be memory-wiped like digital systems? |
| P024.8 | Do human assessors of very large digital objects add insight beyond their tools? |
| P024.9 | How can unilaterally supplied devices be securely installed into a multilateral cross-checking system? |
| P024.10 | Can impostor taps be swapped in before installation, and what custody protocol prevents it? |
Possibly relevant work found by automated search
Found by AI agents searching for this problem's sub-questions, mostly by keyword and citation in the OpenAlex scholarly database and on the web; a second AI pass reviewed each one, and we spot-checked some by hand. They seemed reasonably related, as far as one can tell from outside each paper's field. Some come from neighbouring fields, such as arms control and financial audit, and bear on the problem by analogy. Not reviewed against the status above; a paper listed here does not change that status. 928 of 1052 quotes checked word for word against the paper's text.
For P024.2: Has a protocol been demonstrated by which two parties confirm deletion of data from an attested cluster (e.g. attested key destruction)?
For P024.3: Has a two-party 'either can zeroise' control for a server cluster been built, with an attested zeroisation time?
For P024.4: Has a one-bit output channel from a sealed facility been specified with a bound on extra bits leaked (e.g. via timing)?
For P024.5: What fault/retry rate can be pre-agreed so that deliberate faults (each leaking one bit) stay within an acceptable leakage budget, and what is the benign fault rate of a redundant auditing environment?
For P024.6: How is power supplied to the auditing environment without opening an attack or leak path?
For P024.9: How can unilaterally supplied devices be securely installed into a multilateral cross-checking system?
* Full text not reviewed; more likely to contain errors.
Data
- problems.csv — one row per problem
- problem_status.csv — status, scope and evidence per problem
- problem_sources.csv — the posed-by quote and other documents per problem
- leaves.csv — all sub-questions, including those without a checkable form
- leaf_sources.csv — every source row behind each sub-question