P047 Completeness: showing that declarations cover all compute
Can a state show, to a stated confidence, that it has no compute, stockpiles or facilities outside its declarations?
Whether a state can show that its declarations cover essentially all of its compute, leaving no hidden facilities, stockpiles, parallel infrastructure or exit route, and to what confidence.
Status
Analyzed A proposed procedure only: mutual declaration of existing, in-transit and in-production compute, a supply-chain records audit with inspectors counting final chips, comparison against each side's intelligence, then extension to third nations; no confidence level, no pilot, and the plan itself says unknown compute cannot be brought under verification. Facilities clause: Koopmanschap & Barten's desk analysis finds that with active concealment facilities of 10,000 to 100,000 H100-equivalents cannot be ruled out. Stockpile and secret-production clauses: their rough bound of about a million untraced chips. Exit routes and smuggling: stated limits only (Harack et al., Aarne & Petrie).
"Both the US and China should then be able to carry out an audit and inspection of each other’s declaration, and compare it against their private intelligence efforts (to the extent they properly prepared this) to resolve their uncertainties over how much compute each other have."
Posed by
- Harack et al. (2025), Verification for International AI Governance (Oxford AIGI):
there is no guarantee that the Prover is not maintaining a secret parallel system
(§2.4.1.1, p.56, L3059) (source states the problem but not the quantity the question asks for)
Also raised in: Baker, Kulp, Marks, Brundage & Heim (2025), Verifying International Agreements on AI: Six Layers of Verification for Rules on Large-Scale AI Development and Deployment (RAND WR-A4077-1); Scher & Thiergart (2024), Mechanisms to Verify International Agreements About AI Development (MIRI); Petrie, Aarne, Ammann & Dalrymple (2025), Overview of Flexible Hardware-Enabled Guarantees (flexHEG Part 1); Aarne & Petrie (2025), International Security Applications of Flexible Hardware-Enabled Guarantees (flexHEG Part 3); Shavit (2023), What Does It Take to Catch a Chinchilla? Verifying Rules on Large-Scale Neural Network Training via Compute Monitoring; Rahman (2026), Does Distributed Training Undermine Compute Governance? (MIRI); Dean, Verification Plan (AI 2040); Proofworks, Living Verification Documents.
Sub-questions
| ID | Sub-question | What the sources report |
|---|---|---|
| P047.1 | What share of world AI compute (H100e) sits in sites above stated size cutoffs (e.g. 10K / 100K H100e)? | open |
| P047.2 | What detection probability do intelligence plus inspections give for a stockpile of a stated size of non-mechanism accelerators or consumer GPUs? | implied by a source, not stated as open |
| P047.3 | Is there a fab-monitoring method with a stated probability of detecting undeclared (non-mechanism) AI-chip production at a given volume? | open |
| P047.4 | How many months from a decision to resume non-mechanism production until a frontier-scale cluster is fielded, and is it detectable in that window? | open |
| P047.5 | Does detection probability rising with run duration push evaders to short runs that still exceed thresholds? | implied by a source, not stated as open |
| P047.6 | Can procurement/financial intelligence detect cluster-scale purchases by state actors and non-market economies (as claimed for private firms in market economies)? | implied by a source, not stated as open |
Sub-questions not yet stated precisely enough to check
| ID | Sub-question |
|---|---|
| P047.7 | How can a state credibly demonstrate that no compute or frontier development exists outside declared or hardware-verified sites? |
| P047.8 | How should discovery of one secret data centre update belief in others? |
| P047.9 | Can an IAEA-style programme for spotting and monitoring undeclared data centres be built? |
| P047.10 | Which data centres must register (e.g. a stated power threshold)? |
| P047.11 | Is supplier information and design-information verification adversarially robust enough to supplement other mechanisms? |
| P047.12 | Will parties withhold small (military) data centres from declarations, and can military hardware commitments be verified byte-by-byte elsewhere? |
| P047.13 | Can equipment-based limits or multilateral export controls be made durable against later withdrawal or defection? |
| P047.14 | Does prover-held licensing let a prover exit while retaining full hardware control? |
| P047.15 | Can verification mechanisms deter smuggling of chips beyond the authorities' reach, beyond buyer vetting? |
| P047.16 | Is a list of failure modes (compromised manufacturing, diversion to secret DC, in-place tampering) complete, and which path is cheapest for an adversary? |
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 P047.2: What detection probability do intelligence plus inspections give for a stockpile of a stated size of non-mechanism accelerators or consumer GPUs?
For P047.3: Is there a fab-monitoring method with a stated probability of detecting undeclared (non-mechanism) AI-chip production at a given volume?
For P047.6: Can procurement/financial intelligence detect cluster-scale purchases by state actors and non-market economies (as claimed for private firms in market economies)?
For P047.7: How can a state credibly demonstrate that no compute or frontier development exists outside declared or hardware-verified sites?
For P047.9: Can an IAEA-style programme for spotting and monitoring undeclared data centres be built?
* 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