# AI verification proposal

A proposal built with the Proposal Explorer of the AI Verification Tech Map (https://trustbutveri.fyi/), from its records of 2026-10-08. Interactive version: https://trustbutveri.fyi/explorer/?mechanisms=M-0004,M-0001&implementations=M-0004:I-0003

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 readiness and open flaws. Readiness levels R0 to R4 describe one record's public evidence for its assessed use and are never combined. Definitions: https://trustbutveri.fyi/about/methodology/ (roles, properties and flaws) and https://trustbutveri.fyi/about/readiness/ (readiness levels).

## 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 flaws: critical / significant / minor. The last three columns are the editors' reading of what the verifier sees.

| Mechanism | Readiness | Prover | Attack testing | Hardware | Open flaws | Weights | Inputs and outputs | Training data |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Zero-knowledge proofs of inference / zkLLM | R2 | Adversarial | Analysis | None | 0 / 1 / 1 | hidden | shown | not involved |
| Sampled inference recomputation | R3 | Adversarial | Analysis | None | 0 / 3 / 1 | depends | depends | not involved |

## Claims

No claims chosen.

## Mechanisms

### Zero-knowledge proofs of inference

zkLLM is a GPU-accelerated zero-knowledge proof system that proves a large language model's output came from committed weights without revealing those weights. ([Zero-knowledge proofs of inference](https://trustbutveri.fyi/mechanisms/zk-proofs-of-inference/))

- Assessment: selected implementation [zkLLM](https://trustbutveri.fyi/implementations/zkllm/).
- Readiness: R2 Demonstrated, assessed for proving an output came from committed weights, against a prover who cheats.
- 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 hidden; inputs and outputs shown; training data not involved. zkLLM's stated setting keeps model parameters private while the verifier supplies the prompt and checks the returned output.

### Sampled inference recomputation

A verifier re-runs a random sample of an AI provider's logged queries on a trusted copy of the declared model and checks the outputs match. ([Sampled inference recomputation](https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/))

- Assessment: mechanism family.
- Readiness: R3 In production, assessed for checking untrusted workers' activations against the declared model, prompt and precision.
- 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 depends; inputs and outputs depends; training data not involved. Recomputation needs the weights and sampled requests inside the checking environment. For closed models, the record describes a trusted, confidential environment; disclosure to the verifier depends on that boundary.


## Properties

**In production**

- Sampled inference recomputation: R3 In production, assessed for checking untrusted workers' activations against the declared model, prompt and precision

**Built for an adversarial prover**

- Zero-knowledge proofs of inference
- Sampled inference recomputation

**No new hardware needed**

- Zero-knowledge proofs of inference
- Sampled inference recomputation


## Attack testing

Published attempts to break a system, including those that found failures. Testing history does not establish that open flaws are resolved.

**Testing history**

- Zero-knowledge proofs of inference / zkLLM: Analysis
- Sampled inference recomputation: Analysis


## Limits

**Open significant flaws**

- Reference code is interactive and runs prover and verifier together (open question, in zkLLM; https://trustbutveri.fyi/implementations/zkllm/#flaw-1) [2]. The README states that prover and verifier work "are implemented side-by-side", and that intermediate values written to files are for the prover's reference only. It says an industrial deployment would need to separate the two and apply Fiat–Shamir to make proofs non-interactive. The released code gives a verifier no standalone check.
- Tolerance for numerical noise leaves a covert channel (demonstrated attack, in Sampled inference recomputation; https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/#flaw-1) [9][17][20]. 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.
- Only recorded traffic is checked (theoretical argument, in Sampled inference recomputation; https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/#flaw-2) [9][21]. 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.
- Some inference optimizations are not covered (theoretical argument, in Sampled inference recomputation; https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/#flaw-3) [10][11]. 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.

**Family finding context**

- Context for zkLLM; applicability depends on the finding's scope. The proof covers a fixed-point approximation, not the floating-point model (open question, in Zero-knowledge proofs of inference; https://trustbutveri.fyi/mechanisms/zk-proofs-of-inference/#flaw-1) [1][3][4][5]. Current ZK inference systems prove a quantised version of the network. zkLLM scales values by 2^16 and reports small perplexity changes. Attestable reports quantising matrix multiplications to 8-bit integers while proving other operations in floating point. A verifier therefore learns about the proof-friendly variant, and must separately accept that this variant is the declared model. Trail of Bits built a ResNet-18 backdoor that is dormant in the full-precision model and active after ezkl's quantisation; whether it persists through proving was left for further investigation. A verification system design calls floating-point emulation in ZKPs an open problem.
- Context for zkLLM; applicability depends on the finding's scope. A proof speaks only for the computations that were proven (theoretical argument, in Zero-knowledge proofs of inference; https://trustbutveri.fyi/mechanisms/zk-proofs-of-inference/#flaw-2) [6]. Attestable writes that "a proof of some computation is not a proof of all computation", and that a proof cannot discover a datacenter that was never declared. Proofs of inference do not by themselves show that no other workload ran on the same or other hardware.

  Related mechanism: Proofs of useful work for capacity accounting (R1, not in the proposal). The record names proof-of-work accounting as the kind of compute accounting needed to show that proven inference was the only work done.
- Context for zkLLM; applicability depends on the finding's scope. The model architecture is disclosed (theoretical argument, in Zero-knowledge proofs of inference; https://trustbutveri.fyi/mechanisms/zk-proofs-of-inference/#flaw-3) [1][7]. ZKML "requires that the model architecture (but not weights) is revealed", and zkLLM assumes a publicly known model structure. Architecture can be commercially sensitive.
- Context for zkLLM; applicability depends on the finding's scope. Proofs do not bind computational effort (Hollow-LLM) (demonstrated attack, in Zero-knowledge proofs of inference; https://trustbutveri.fyi/mechanisms/zk-proofs-of-inference/#flaw-4) [8]. Researchers at the University of Southern California show that a proof of inference certifies that an output is consistent with committed weights under the declared architecture, but not how much computation produced it. In their Hollow-LLM attack, a provider keeps the declared architecture and parameter count but commits to "ghost weights". Some layers pass their inputs through unchanged, and wide layers carry the signal in a small subspace, so a much smaller inner model does the real work. The ghost weights satisfy the verification circuit and yield valid proofs.

  The authors ran the attack with the proof procedure of zkGPT, a separate ZK inference system, on a 6-layer, 512-dimensional transformer declared as up to 12 layers and 1,024 dimensions. Outputs were identical to the inner model's, and serving cost stayed at the inner model's level. An honest model of the declared size cost 2.4 times as much to prefill and 3.1 times as much to decode. Proving cost still grew with the declared architecture.

  The authors note that results may be served before any proof, with the provider building the witness only when a call is selected for audit. They describe their constructions as "compatible with state-of-the-art zkLLM pipelines", and state that the attack does not imply a flaw in the proof system itself. They propose challenge-based audits and ablation tests, which raise the cost of cheating but give no guarantee.

**Open minor flaws**

- Proves a fixed-point approximation of a publicly known architecture (theoretical argument, in zkLLM; https://trustbutveri.fyi/implementations/zkllm/#flaw-2) [1]. The prover's model must have a "publicly known structure". Tensors are discretised by scaling and rounding. The authors report perplexity changes of 0.008 to 0.09 on C4. The proof covers the quantised computation.
- Mixed hardware widens the honest baseline (open question, in Sampled inference recomputation; https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/#flaw-4) [10]. 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.


## Possible additions

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

- **Deterministic and bit-exact inference** (R3 In production, assessed for reproducing open-model inference from receipts in Gensyn's information-market service)
  - Bears on the open significant flaw "Tolerance for numerical noise leaves a covert channel" in Sampled inference recomputation. Bit-exact inference would remove the numerical tolerance if exact replay can be deployed with the required weights and configuration.
- **Network taps and certifiers** (R1 Proposed, assessed for committing a complete record of cluster traffic, so declared inference can be checked)
  - Bears on the open significant flaw "Only recorded traffic is checked" in Sampled inference recomputation. 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.
  - Sampled inference recomputation waits on it: In tap-based retrofit designs, recording all inference traffic needs network taps and recomputation servers that can ingest it, in the worst case one recomputation-server network interface per inference front-end interface.


## Dependencies

**Blockers**

- Zero-knowledge proofs of inference: Proving takes about 13 minutes (803 seconds) of A100 time per 2,048-token forward pass at 13B parameters, plus a one-time weight commitment of 16 to 21 minutes. (performance & compatibility) [1]
- Zero-knowledge proofs of inference: The repository was archived on 10 July 2025 and the author states there is no plan for upgrades or maintenance. (performance & compatibility) [2]
- Zero-knowledge proofs of inference: No security audit of the code has been carried out. (adversarial validation) [2]
- Sampled inference recomputation: In tap-based retrofit designs, recording all inference traffic needs network taps and recomputation servers that can ingest it, in the worst case one recomputation-server network interface per inference front-end interface. (coverage & hidden compute; waits on Network taps and certifiers) [16][22]
- Sampled inference recomputation: In retrofit designs, the recomputation server must sit inside the prover's data centre, possibly under the prover's physical control, and still be protected from a compromised provider, which Amodo rates 'not on track'. (hardware trust) [9][16]
- Sampled inference recomputation: No independent red-team of a recomputation consistency check has been published (the one independent attack study targets the weight-exfiltration bound), and Amodo rates recomputation red-teaming 'not started'. (adversarial validation) [16][17]
- Sampled inference recomputation: Tolerance-based checks need calibration on trusted hardware and exact knowledge of the provider's sampling procedure, and in one prototype a sampling-implementation mismatch produced large spurious differences. (performance & compatibility) [10][15]
- Sampled inference recomputation: The verifier needs the model weights, so checking a closed-weights model requires a trusted, confidential recomputation environment, which the retrofit designs place inside the prover's facility. (privacy & leakage) [5][10][21]


## What the verifier sees

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

## Implementations

- Zero-knowledge proofs of inference: [Attestable zero-knowledge inference prover](https://trustbutveri.fyi/implementations/attestable-zk-inference/) (R1, product); [EZKL](https://trustbutveri.fyi/implementations/ezkl/) (R2, product); [Low-trust AI compute verification system overview](https://trustbutveri.fyi/implementations/low-trust-compute-verification-system-overview/) (R1, proposed architecture); [zkLLM](https://trustbutveri.fyi/implementations/zkllm/) (R2, research prototype)
- 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. zkLLM: Zero Knowledge Proofs for Large Language Models, H. Sun et al. (2024). https://doi.org/10.1145/3658644.3670334
2. zkllm-ccs2024: code for zkLLM: Zero Knowledge Proofs for Large Language Models, H. Sun (2024). https://github.com/jvhs0706/zkllm-ccs2024
3. Proving LLMs at Scale, Attestable (2026). https://attestable.com/blog/proving-llms-scale
4. Zkonduit EZKL Security Assessment, F. Casal et al. (2025). https://github.com/trailofbits/publications/blob/master/reviews/2025-03-zkonduit-ezkl-securityreview.pdf
5. A System Overview for Near-Term, Low-Trust AI Compute Verification, N. Cankaya (2026). https://intelligence.org/wp-content/uploads/2026/06/A-system-overview-for-near-term-low-trust-AI-compute-verification.pdf
6. Pacing AI Requires Proof, Attestable (2026). https://attestable.com/blog/pacing-ai-requires-proof
7. ZKML: An Optimizing System for ML Inference in Zero-Knowledge Proofs, B.-J. Chen et al. (2024). https://doi.org/10.1145/3627703.3650088
8. Hollow-LLM Attack: Computationally Trivial Weights in Zero-Knowledge Verification of LLM Inference, C. Gong et al. (2026). https://arxiv.org/abs/2607.28884
9. Verifying LLM Inference to Detect Model Weight Exfiltration, R. Rinberg et al. (2025). https://arxiv.org/abs/2511.02620
10. DiFR: Inference Verification Despite Nondeterminism, A. Karvonen et al. (2025). https://arxiv.org/abs/2511.20621
11. TOPLOC: A Locality Sensitive Hashing Scheme for Trustless Verifiable Inference, J. M. Ong et al. (2025). https://proceedings.mlr.press/v267/ong25a.html
12. PrimeIntellect-ai/toploc (GitHub repository), Prime Intellect (2025). https://github.com/PrimeIntellect-ai/toploc
13. INTELLECT-2: A Reasoning Model Trained Through Globally Decentralized Reinforcement Learning, Prime Intellect Team et al. (2025). https://arxiv.org/abs/2505.07291
14. adamkarvonen/difr (GitHub repository), A. Karvonen (2025). https://github.com/adamkarvonen/difr
15. Scaling Recomputation Inference Verification, Amodo Design (2026). https://amododesign.com/notes/2026-09-02-scaling-recomputation-inference-verification/
16. AI 2040 Plan A — Verification SITREP, Amodo Design (2026). https://amododesign.com/ai-verification/plan-a-sitrep/
17. Adversarial Entropy Inflation Against Gumbel-Based Inference Verification, N. Kezins (2026). https://arxiv.org/abs/2608.23375
18. SYNTHETIC-2 Release: Four Million Collaboratively Generated Reasoning Traces, Prime Intellect (2025). https://www.primeintellect.ai/blog/synthetic-2-release
19. An Inference Verification Prototype — Stage 1, Amodo Design (2026). https://amododesign.com/notes/2026-06-29-inference-verification-prototype/
20. Bit-Exact AI Inference Verification Without Performance Tradeoffs, N. Cankaya (2026). https://arxiv.org/abs/2606.00279
21. Example Schemes for Verifying High-Stakes AI Agreements, Amodo Design (2026). https://amododesign.com/notes/2026-06-23-verification-algorithms/
22. Verification Plan, R. Dean (2026). https://ai-2040.com/supplements/verification-plan