P017 Legacy installed base and retrofit

Which verification functions can be retrofitted onto the existing installed base of accelerators by firmware alone or an external module, and what fraction of installed compute cannot take them?

Retrofitting verification functions onto the legacy installed base.

Terms: H100 / A100: Nvidia's current and prior-generation data-centre GPUs, used here to stand for the large existing installed base of accelerators.

Status

Analyzed A desk analysis drawing on industry conversations: which governance mechanisms the installed base could take by firmware alone (the H100's security module, TEE and secure boot are noted as present), with timeline estimates by threat context (months for firmware-only, 18 months to 4 years for silicon changes, several years for tamper-proofing) and the caveat that existing secure-boot implementations likely carry unpatchable flaws. External-module retrofit is a design option in flexHEG Part 2 (capacitor-based disablement 'a good option for retrofitting'; antifuses unsuitable). No held document demonstrates a retrofit or states what fraction of installed compute cannot take one; Shavit (2023) notes many sold chips may be neither retrofittable nor locatable. Brass & Aarne (2024), cited by Tee & Happel for delay-based location verification on existing confidential-computing keys, is not held.

For minimally adversarial contexts the current level of hardware security likely would be sufficient, and thus many mechanisms could be implemented as firmware updates. This would take a few months. Some mechanisms may not be possible to implement this way on current hardware, in which case silicon-level changes would be required, and the time to implement them would increase to between 18 months and 4 years.
Aarne, Fist & Withers (2024), Secure, Governable Chips (CNAS) (2024-01), L2199–2206

Posed by

Also raised in: 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); Wasil, Reed, Miller & Barnett (2024), Verification Methods for International AI Agreements; AI Security Forum, Requests for Discussion.

Sub-questions

IDSub-questionWhat the sources report
P017.1Which verification functions (licence enforcement, cluster attestation, secure timer, delay-based location, signed-firmware enforcement) can be added to the installed H100/A100 base by firmware alone, and what fraction of installed FLOP/s lacks them?open

Sub-questions not yet stated precisely enough to check

IDSub-question
P017.2What fraction of a deployed GPU fleet could be retrofitted with an external guarantee module, at what cost per unit?

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 P017.1: Which verification functions (licence enforcement, cluster attestation, secure timer, delay-based location, signed-firmware enforcement) can be added to the installed H100/A100 base by firmware alone, and what fraction of installed FLOP/s lacks them?

Data