# AI verification proposal

A proposal built with the Proposal Explorer of the AI Verification Tech Map (https://trustbutveri.fyi/), from its records of 2026-10-09. Interactive version: https://trustbutveri.fyi/explorer/?mechanisms=M-0020,M-0001&implementations=M-0001:I-0002

How to read it: a claim is something one party wants to verify about another's AI hardware or software. A mechanism is a general technique for verifying claims; it is "aimed at" a claim when that is its direct purpose, and "supporting" when it contributes without being aimed at it. A claim is addressed when a mechanism in the proposal is aimed at it and is not excluded by the filters; addressed does not mean verified, so check that mechanism's development status, security evidence and findings. Definitions: https://trustbutveri.fyi/about/methodology/ (roles, properties and findings) and https://trustbutveri.fyi/about/readiness/ (development status).

## Filters

Filters apply to mechanisms only and describe the setting the proposal is for.

None set. Every mechanism on the map was available.

## Overview

One row per mechanism, read from its record. Open failures: critical / significant / minor. The last three columns are the editors' reading of what the verifier sees. Findings are grouped as known failures, scope limitations and open questions. Only known failures count as failures. Counts are an inventory of published findings, not a risk score.

| Mechanism | Development | Security evidence | Prover | Attack testing | Hardware | Open failures | Weights | Inputs and outputs | Training data |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Remote detection of data centres | Proposed | Published security analysis | Adversarial | Analysis | None | 0 / 0 / 0 | not involved | not involved | not involved |
| Sampled inference recomputation / DiFR (Divergence From Reference) | Research demonstration | Published security analysis | Adversarial | Analysis | None | 0 / 1 / 1 | unspecified | unspecified | unspecified |

## Claims

No claims chosen.

## Mechanisms

### Remote detection of data centres

Remote detection locates large data centres and estimates their power capacity without site access, using satellite imagery, heat signatures and public records such as permits. ([Remote detection of data centres](https://trustbutveri.fyi/mechanisms/remote-detection-of-data-centres/))

- Assessment: mechanism family.
- Development: Proposed (legacy code R1), assessed for finding undeclared data centres above an agreed compute threshold.
- Security evidence: Published security analysis. Independent evaluation: unassessed. Formal proof: unassessed. Deployment assurance: unassessed.
- Claims in this proposal: none of them.
- Threat model: adversarial prover. Hardware: none. Prover cooperation: not required. Attack testing: analysis. Category: Remote & side-channel sensing.
- What the verifier sees: model weights not involved; inputs and outputs not involved; training data not involved. Works from outside the facility; it does not handle model data.

### Sampled inference recomputation

DiFR checks that an inference provider ran its declared model by comparing output tokens or activations with a trusted re-run using the same random seed. ([Sampled inference recomputation](https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/))

- Assessment: selected implementation [DiFR (Divergence From Reference)](https://trustbutveri.fyi/implementations/difr/).
- Development: Research demonstration (legacy code R2), assessed for checking that outputs match the declared model, precision and sampling settings.
- Security evidence: Published security analysis. Independent evaluation: unassessed. Formal proof: unassessed. Deployment assurance: unassessed.
- Claims in this proposal: none of them.
- Threat model: adversarial prover. Hardware: none. Prover cooperation: required. Attack testing: analysis. Category: Cryptographic & computational.
- What the verifier sees: model weights unspecified; inputs and outputs unspecified; training data unspecified. This Explorer has no asset-specific exposure assessment for this implementation. Check its source and deployment assumptions.


## Properties

**Built for an adversarial prover**

- Remote detection of data centres
- Sampled inference recomputation

**No new hardware needed**

- Remote detection of data centres
- Sampled inference recomputation


## Attack testing

Attack testing records published testing for this use. It does not by itself show independent review, a formal proof or that a deployed system is secure.

**Testing history**

- Remote detection of data centres: Analysis
- Sampled inference recomputation / DiFR (Divergence From Reference): Analysis


## Limits

**Open significant failures**

- Statistical tolerance leaves a covert channel (known failure, demonstrated attack, in DiFR (Divergence From Reference); https://trustbutveri.fyi/implementations/difr/evidence/flaws/1/) [7][12][14]. Statistical schemes can put an upper bound on an adversary's covert bandwidth, but cannot close it. In the companion exfiltration study, the detector cut exfiltratable information to under 0.5% under benign prompt traffic. It did not cut it to zero. Its authors called the channel's size under adversarial prompts an open empirical question. An independent study attacked that detector, which uses the same Gumbel-margin statistic. An adversary who controls the prompts roughly doubled the bits leaked per token. Across six models, this cut the slowdown from 146–254 times under benign prompts to 60–118 times. The attack widens the exfiltration bound. It does not target DiFR's check that outputs match the declared configuration.

**Family finding context**

- Context for DiFR (Divergence From Reference). Findings from the mechanism family appear here as context. They apply to an implementation only when its own record lists them, under the conditions stated there. Tolerance for numerical noise leaves a covert channel (known failure, demonstrated attack, in Sampled inference recomputation; https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/evidence/flaws/1/) [7][12][14]. Schemes that accept approximate matches can put an upper bound on an adversary's covert bandwidth, but they cannot close the channel. The weight-exfiltration detector cut exfiltratable information to under 0.5%, not to zero, on a 30-billion-parameter mixture-of-experts model under benign prompt traffic. Its authors called the channel's size under adversarial prompts an open empirical question. An independent study showed that an adversary who controls the prompts roughly doubles the bits leaked per token. Across six models, that cut the slowdown from 146–254 times under benign prompts to 60–118 times. The attack widens the exfiltration bound. It does not target the check that outputs match the declared model.

  Related mechanism: Deterministic and bit-exact inference (R3, not in the proposal). Bit-exact inference would remove the numerical tolerance if exact replay can be deployed with the required weights and configuration.
- Context for DiFR (Divergence From Reference). Findings from the mechanism family appear here as context. They apply to an implementation only when its own record lists them, under the conditions stated there. Only recorded traffic is checked (scope limitation, theoretical argument, in Sampled inference recomputation; https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/evidence/flaws/2/) [7][15]. Recomputation checks that recorded, declared workloads are correct. It cannot show that the record is complete. The published schemes do not cover hidden workloads run on the same compute, or substituted work. Rinberg et al. say their exfiltration-detection scheme cannot stand alone.

  Related mechanism: Network taps and certifiers (R1, not in the proposal). Taps copy and hash all traffic on the monitored links, which bears on whether the traffic record is complete. They do not show what else ran on the same chips.
- Context for DiFR (Divergence From Reference). Findings from the mechanism family appear here as context. They apply to an implementation only when its own record lists them, under the conditions stated there. Some inference optimizations are not covered (known failure, theoretical argument, in Sampled inference recomputation; https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/evidence/flaws/3/) [6][16]. TOPLOC's authors state that it cannot detect speculative decoding in which a cheaper model does the decoding. They did not test whether it distinguishes types of key-value (KV) cache compression. DiFR was evaluated only on sampling from a single model. Its authors sketch an extension to one speculative-decoding algorithm but do not test it.
- Context for DiFR (Divergence From Reference). Findings from the mechanism family appear here as context. They apply to an implementation only when its own record lists them, under the conditions stated there. Mixed hardware widens the honest baseline (known failure, open question, in Sampled inference recomputation; https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/evidence/flaws/4/) [6]. When honest reference runs span different GPU types, the spread of benign scores grows. In DiFR's tests on Qwen3-30B-A3B, pooling A100 and H200 runs left Token-DiFR unable to separate the two smallest tested changes, a temperature of 1.1 instead of 1.0 and a simulated top-2 sampling bug, at the target false-positive rate, while cross-entropy separated them. Matched provider and verifier environments, or pooling that weights rare large deviations, restored detection.

**Open minor failures**

- Mixed hardware widens the honest baseline (known failure, open question, in DiFR (Divergence From Reference); https://trustbutveri.fyi/implementations/difr/evidence/flaws/2/) [6]. For Qwen3-30B-A3B, pooling honest runs across A100 and H200 GPUs and parallelism setups broadened the honest score distribution. Token-DiFR then failed to separate the two smallest tested changes, a temperature of 1.1 instead of 1.0 and a simulated top-2 sampling bug, at the target false-positive rate, while cross-entropy did. The authors report that matched provider and verifier environments, or pooling that weights rare large deviations, restore detection.

**Scope limitations**

- Facilities can be disguised or hidden (scope limitation, theoretical argument, in Remote detection of data centres; https://trustbutveri.fyi/mechanisms/remote-detection-of-data-centres/evidence/flaws/1/) [1]. Halstead and Larsen discuss two ways to hide a facility. One is to disguise it as a legitimate industrial site. The other is to build it underground, with cooling that avoids visible heat plumes. They note that the underground option requires bespoke engineering.
- Small sites may not be detectable (scope limitation, theoretical argument, in Remote detection of data centres; https://trustbutveri.fyi/mechanisms/remote-detection-of-data-centres/evidence/flaws/2/) [1][3]. Halstead and Larsen conclude that a sufficiently small covert project could not be ruled out with confidence. In their estimates, the chance of detection is lower for smaller sites. Krawec notes that small data centres in existing buildings may lack the distinctive features of large facilities.

  Related mechanism: Chip registries and manufacturing records (R1, not in the proposal). Accounts for chips from the fab onwards, which does not depend on a site being visible.

**Open questions**

- Search for unknown sites is undemonstrated (open question, open question, in Remote detection of data centres; https://trustbutveri.fyi/mechanisms/remote-detection-of-data-centres/evidence/flaws/3/) [3]. Krawec reports that telling data centres apart from other industrial facilities systematically is difficult. Automating detection would need large amounts of training imagery and a purpose-trained model. In Krawec's words, automated data-centre detection "remains primarily conceptual at present".
- Speculative decoding and multi-model sampling not evaluated (open question, open question, in DiFR (Divergence From Reference); https://trustbutveri.fyi/implementations/difr/evidence/flaws/3/) [6]. The algorithms and experiments cover sampling from a single LLM. Speculative decoding was not evaluated. The authors sketch an extension to one speculative-decoding algorithm, without experiments. They note that other variants would need modified verification and extra metadata.

**Not yet demonstrated**

- Remote detection of data centres: Proposed (legacy code R1), assessed for finding undeclared data centres above an agreed compute threshold


## Possible additions

Mechanisms on the map, not in the proposal, that the records connect to an unaddressed or partly addressed claim, an open failure or a dependency. Pointers, not recommendations: each brings its own readiness level and findings, and none is claimed to close a failure.

None found.



## Dependencies

**Blockers**

- Remote detection of data centres: Wide-area, automated detection of data centres is not yet practical and needs large training datasets. (coverage & hidden compute) [3]
- Remote detection of data centres: No measured detection or false-alarm rates for finding undeclared facilities have been published. (adversarial validation) [1][3]
- Remote detection of data centres: Recent high-resolution imagery is costly, is limited by weather and needs trained analysts. (access & governance) [3]
- Sampled inference recomputation: The verifier needs the model weights, so outsiders cannot use the method to verify providers of closed-weights models. (privacy & leakage) [6]
- Sampled inference recomputation: The verifier must know and match the provider's sampling procedure, and in one prototype a sampling mismatch in a newer vLLM version produced large spurious logit differences. (performance & compatibility) [6][9]
- Sampled inference recomputation: No independent red-team of DiFR's consistency check has been published, Amodo rates recomputation red-teaming 'not started', and the one independent attack study targets an exfiltration detector built on the same statistic. (adversarial validation) [11][12]


## What the verifier sees

- Model weights: shown by none; depends on the design for none; hidden by none; not involved in Remote detection of data centres; unspecified for Sampled inference recomputation.
- Inputs and outputs: shown by none; depends on the design for none; hidden by none; not involved in Remote detection of data centres; unspecified for Sampled inference recomputation.
- Training data: shown by none; depends on the design for none; hidden by none; not involved in Remote detection of data centres; unspecified for Sampled inference recomputation.

## Implementations

- Remote detection of data centres: none on the map
- Sampled inference recomputation: [AI 2040 inference-only verification stack](https://trustbutveri.fyi/implementations/ai-2040-inference-only-verification-plan/) (R1, proposed architecture); [DiFR (Divergence From Reference)](https://trustbutveri.fyi/implementations/difr/) (R2, research prototype); [Low-trust AI compute verification system overview](https://trustbutveri.fyi/implementations/low-trust-compute-verification-system-overview/) (R1, proposed architecture); [SASH confidential network logger](https://trustbutveri.fyi/implementations/sash-confidential-network-logger/) (R1, research prototype); [TOPLOC](https://trustbutveri.fyi/implementations/toploc/) (R3, open-source project)

## Sources

1. Covert AI Projects, B. Halstead & T. Larsen (2026). https://ai-2040.com/supplements/covert-ai-projects
2. Verifying International Agreements on AI: Six Layers of Verification for Rules on Large-Scale AI Development and Deployment, M. Baker et al. (2025). https://www.rand.org/pubs/working_papers/WRA4077-1.html
3. Tracking Hyperscale AI Data Center Growth with Satellite Imagery, C. Krawec (2026). https://fas.org/publication/tracking-hyperscale/
4. Introducing the Frontier Data Centers Hub, Epoch AI (2025). https://epoch.ai/latest/introducing-the-frontier-data-centers-hub
5. AI Data Centers Documentation – Methodology, Epoch AI (2026). https://epoch.ai/data/data-centers-documentation/methodology
6. DiFR: Inference Verification Despite Nondeterminism, A. Karvonen et al. (2025). https://arxiv.org/abs/2511.20621
7. Verifying LLM Inference to Detect Model Weight Exfiltration, R. Rinberg et al. (2025). https://arxiv.org/abs/2511.02620
8. adamkarvonen/difr (GitHub repository), A. Karvonen (2025). https://github.com/adamkarvonen/difr
9. Scaling Recomputation Inference Verification, Amodo Design (2026). https://amododesign.com/notes/2026-09-02-scaling-recomputation-inference-verification/
10. Amodo-Design/Inference-Recomputation-Prototype (GitHub repository), Amodo Design (2026). https://github.com/Amodo-Design/Inference-Recomputation-Prototype
11. AI 2040 Plan A — Verification SITREP, Amodo Design (2026). https://amododesign.com/ai-verification/plan-a-sitrep/
12. Adversarial Entropy Inflation Against Gumbel-Based Inference Verification, N. Kezins (2026). https://arxiv.org/abs/2608.23375
13. An Inference Verification Prototype — Stage 1, Amodo Design (2026). https://amododesign.com/notes/2026-06-29-inference-verification-prototype/
14. Bit-Exact AI Inference Verification Without Performance Tradeoffs, N. Cankaya (2026). https://arxiv.org/abs/2606.00279
15. Example Schemes for Verifying High-Stakes AI Agreements, Amodo Design (2026). https://amododesign.com/notes/2026-06-23-verification-algorithms/
16. TOPLOC: A Locality Sensitive Hashing Scheme for Trustless Verifiable Inference, J. M. Ong et al. (2025). https://proceedings.mlr.press/v267/ong25a.html