P009 Production oversight of chips and mechanism components
What institutional oversight - monitoring, inspection standards, independent testing and verified destruction - covers the production of chips and of the components that verification mechanisms depend on?
Institutional oversight of chip and mechanism-component production: monitoring regimes, inspection standards and capacity, independent testing, verified destruction of failed units, tool-use controls, and retroactively establishing trust in already-produced chips.
Terms: p: the fraction of devices sampled for imaging. d: the chance that imaging catches a compromised device it inspects. n: the number of compromised devices.
Status
Analyzed A design for an international body's oversight of guarantee-processor production: manufacturing equipment locked to the approved design, independent testing by a party other than the manufacturer, continuous camera and on-site staff oversight, tamper-evident shipping, extensive testing and imaging of a 1/1000 sample (the authors' own arithmetic gives a 63% chance of catching a 1,000-device compromised batch), verifiable destruction of failed devices, published processes and authorised penetration tests. Nothing instituted or piloted. Oversight of accelerator fabs themselves is only an analogy to nuclear-fuel monitoring (Shavit 2023); retroactive trust in already-produced chips is a proposal to sample and destructively reverse-engineer existing components (flexHEG Part 1). No held document reports any oversight regime in operation.
Each guarantee processor should be subjected to automated, independent testing. … the testing should be done by a different party than the one that manufactured it, ideally either by the international body, several member states, or both.
Posed by
- Aarne, Fist & Withers (2024), Secure, Governable Chips (CNAS):
regulators will need confidence that these mechanisms have not been compromised by untrusted firms or insider attacks during fabrication and packaging.
(App B, Supporting Ecosystem, p30, L2884–2886)
Also raised in: 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; Ilhan, Withers, Gietz & Harack (2026), Verifiable Semiconductor Manufacturing (Oxford AIGI); Cankaya (2025), TSMC Most Definitely Has a Golden Record of All AI Chips It Made; Cankaya (2026), Compute Verification FAQs; Wasil, Reed, Miller & Barnett (2024), Verification Methods for International AI Agreements; Proofworks, Living Verification Documents.
Sub-questions
| ID | Sub-question | What the sources report |
|---|---|---|
| P009.1 | In sample-based audits, what fraction of chip-owner directory entries match the chip found at the declared location after resales and damage? | implied by a source, not stated as open |
| P009.2 | Is there a field-tested method for verifying a registered lithography tool stays on its registered production line? | open |
| P009.3 | Has any lithography or packaging tool enforced a design lock requiring external authorization? | open |
| P009.4 | Is there a supply-chain monitoring protocol for mechanism components (enclosures, guarantee processors) with a quantified probability of catching a compromised batch, including insider compromise at fab and packaging? | open |
| P009.5 | At sampling rate p with imaging detection probability d<1, what is the catch probability for n compromised devices? | partly answered |
| P009.6 | How many days does heavy inspection of a stated number of dies per lot add to fab-to-deployment time, and does it deter? | open |
| P009.7 | Is there a protocol for retroactively establishing trust in already-produced chips, with confidence as a function of sample size? | open |
| P009.8 | Is there a neutral trusted manufacturer candidate for a US–China scenario? | implied by a source, not stated as open |
Sub-questions not yet stated precisely enough to check
| ID | Sub-question |
|---|---|
| P009.9 | Is continuous joint video/portal monitoring of fabs and facilities sufficient, as in nuclear fuel production oversight? |
| P009.10 | Is multinational supply-chain cooperation for global chip accounting logistically and politically achievable? |
| P009.11 | Can distributor sampling verify that consumer GPUs were sold to ordinary consumers? |
| P009.12 | How should facilities producing key mechanism components be separated and overseen? |
| P009.13 | What fab/data-centre inspection standards are needed, given complexity demands specialized expertise? |
| P009.14 | Must device testing be done by a party other than the manufacturer, and how? |
| P009.15 | Has production oversight been subjected to authorized penetration tests (including by intelligence agencies)? |
| P009.16 | How are failed devices verifiably destroyed? |
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 P009.1: In sample-based audits, what fraction of chip-owner directory entries match the chip found at the declared location after resales and damage?
For P009.2: Is there a field-tested method for verifying a registered lithography tool stays on its registered production line?
For P009.4: Is there a supply-chain monitoring protocol for mechanism components (enclosures, guarantee processors) with a quantified probability of catching a compromised batch, including insider compromise at fab and packaging?
For P009.5: At sampling rate p with imaging detection probability d<1, what is the catch probability for n compromised devices?
For P009.8: Is there a neutral trusted manufacturer candidate for a US–China scenario?
* 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