P002 Tamper-respondent enclosures
Can a tamper-respondent enclosure for a high-power accelerator pass liquid cooling and full interconnect while erasing its keys on intrusion, at a low false-trigger rate and a hard-to-defeat threshold?
Tamper-respondent enclosures for accelerators: red-teams, cooling, zeroisation, false triggers.
Terms: FIPS 140 Level 4: the highest tamper-resistance tier in the US/Canada standard for cryptographic hardware, requiring an enclosure that detects and responds to physical intrusion.
Status
Analyzed Design options only, for enclosures around accelerators dissipating up to 1,200 W: active sensing by capacitance, resistance mesh, radio response or distributed sensors; response by key deletion or permanent disablement (capacitor discharge, antifuses, nanothermite); cooling pass-through by a thin boundary under the cold plate, a contained liquid loop with a boundary heat exchanger, or metalised foam for air; enclosure sizes from chip to rack. Nothing built at this scale, no red team, no false-trigger rate; the authors judge that nation-state attackers can likely compromise the best current enclosures and that the per-device cost of circumvention is hard to estimate. The sensing prototypes the section cites are not held. Cross-references, not evidence for this headline: Immler et al. (2019) built and attacked a PCB-scale PUF cover (enclosure PUFs problem); Amodo's tray-level sensor retrofit with power-down response (seals and sensor suites problem) is not an enclosure.
Secure enclosures can ideally be made compatible with both air and direct to chip liquid cooling … To permit cooling, the secure enclosure boundary can either be sandwiched between the chip face and cooling pad, or it can contain both the chip and the cooling pad. If it is sandwiched beneath the cooling pad, it has to be very thin and very thermally conductive.
Posed by
- Reuel, Bucknall et al. (2025), Open Problems in Technical AI Governance, compute sections:
demanding cooling requirements and high-bandwidth interconnect between chips pose a challenge due to the need to bridge the interior and exterior of the tamper-proof enclosure
(§6.2.2, p.35, L1936–1937 (Q76, L1877–1878)) (source states the problem but not the quantity the question asks for)
Also raised in: Harack et al. (2025), Verification for International AI Governance (Oxford AIGI); 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); Petrie & Aarne (2025), Technical Options for Flexible Hardware-Enabled Guarantees (flexHEG Part 2); Aarne & Petrie (2025), International Security Applications of Flexible Hardware-Enabled Guarantees (flexHEG Part 3); Aarne, Fist & Withers (2024), Secure, Governable Chips (CNAS); Kulp et al. (2024), Hardware-Enabled Governance Mechanisms (RAND WR-A3056-1); AI Security Forum, Requests for Discussion.
Sub-questions
| ID | Sub-question | What the sources report |
|---|---|---|
| P002.1 | Has a retrofittable tamper-evident server or rack enclosure passed an independent red-team with a stated detection rate / time-to-defeat? | open |
| P002.2 | What time-to-defeat / per-device cost does the best available accelerator enclosure (e.g. FIPS 140 L4-class mesh) show under a state-resourced red team? | open |
| P002.3 | Has a tamper-respondent enclosure or durable package been built for a ≥700–1,200 W accelerator that passes liquid cooling and full-bandwidth interconnect, and what thermal penalty (K/W) does it add? | open |
| P002.4 | Has a tamper-respondent design zeroised keys (and optionally bricked the chip) within a stated time on intrusion, with a stated false-trigger rate in data-centre operation? | partly answered |
| P002.5 | Has antifuse-based (or other) permanent disablement been shown on a production-class AI die? | open |
| P002.6 | What is the false-positive rate per device-year of mesh tamper sensors in data-centre conditions (lock vs destroy)? | implied by a source, not stated as open |
| P002.7 | What battery capacity powers enclosure tamper sensing for the chip's lifetime, and does it fit the form factor? | open |
| P002.8 | Does radio-response enclosure sensing detect non-metallic (dielectric, laser, FIB) intrusion? | open |
| P002.9 | Does a published certification standard for tamper protection of AI accelerators exist? | open |
Sub-questions not yet stated precisely enough to check
| ID | Sub-question |
|---|---|
| P002.10 | How should a device communicate that tampering occurred (e.g. 'all OK' heartbeat)? |
| P002.11 | Can tamper-protected hardware be repaired / RMA'd without a tamper reset that opens an attack path? |
| P002.12 | Can nested / two-layer enclosures give bilateral trust in shared comparison logic? |
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 P002.1: Has a retrofittable tamper-evident server or rack enclosure passed an independent red-team with a stated detection rate / time-to-defeat?
For P002.2: What time-to-defeat / per-device cost does the best available accelerator enclosure (e.g. FIPS 140 L4-class mesh) show under a state-resourced red team?
For P002.4: Has a tamper-respondent design zeroised keys (and optionally bricked the chip) within a stated time on intrusion, with a stated false-trigger rate in data-centre operation?
For P002.5: Has antifuse-based (or other) permanent disablement been shown on a production-class AI die?
For P002.7: What battery capacity powers enclosure tamper sensing for the chip's lifetime, and does it fit the form factor?
* 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