# 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-0001,M-0014&implementations=M-0001:I-0001

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 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Sampled inference recomputation / TOPLOC | R3 | Adversarial | Analysis | None | 0 / 4 / 0 | unspecified | unspecified | unspecified |
| Bandwidth limits and compartmentalization | R2 | Adversarial | Analysis | Retrofit device | 0 / 5 / 0 | not involved | not involved | not involved |

## Claims

No claims chosen.

## Mechanisms

### Sampled inference recomputation

TOPLOC is a hashing scheme from Prime Intellect that lets a verifier check whether an inference provider ran the model, prompt and precision it claims. ([Sampled inference recomputation](https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/))

- Assessment: selected implementation [TOPLOC](https://trustbutveri.fyi/implementations/toploc/).
- Readiness: R3 In production, assessed for checking that untrusted providers used the claimed 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 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.

### Bandwidth limits and compartmentalization

Capping or removing network links between groups of accelerators, so that serving within each group still works but large training across groups becomes far slower. ([Bandwidth limits and compartmentalization](https://trustbutveri.fyi/mechanisms/bandwidth-limits-and-compartmentalization/))

- Assessment: mechanism family.
- Readiness: R2 Demonstrated, assessed for monitoring inter-node traffic with operator-run software on four GPUs.
- Claims in this proposal: none of them.
- Threat model: adversarial prover. Hardware: retrofit device. Prover cooperation: required. Attack testing: analysis. Category: Isolation & system architectures.
- What the verifier sees: model weights not involved; inputs and outputs not involved; training data not involved. Caps traffic between groups of chips; it does not read the traffic's content.


## Properties

**In production**

- Sampled inference recomputation: R3 In production, assessed for checking that untrusted providers used the claimed model, prompt and precision

**Built for an adversarial prover**

- Sampled inference recomputation
- Bandwidth limits and compartmentalization

**No new hardware needed**

- 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**

- Sampled inference recomputation / TOPLOC: Analysis
- Bandwidth limits and compartmentalization: Analysis


## Limits

**Open significant flaws**

- Speculative decoding goes undetected (theoretical argument, in TOPLOC; https://trustbutveri.fyi/implementations/toploc/#flaw-1) [1]. The TOPLOC authors state that it cannot detect speculative decoding. In speculative decoding, a provider decodes with a cheaper model and uses the larger model only for prefill.
- Last-layer activations could be spoofed (open question, in TOPLOC; https://trustbutveri.fyi/implementations/toploc/#flaw-2) [1]. The TOPLOC authors name spoofing of the last hidden layer's activations as a potential attack. A provider could do this by pruning intermediate layers or by using a smaller model.
- Subtle modifications are harder to detect (open question, in TOPLOC; https://trustbutveri.fyi/implementations/toploc/#flaw-3) [1]. The TOPLOC authors state that large changes to the model or prompt are straightforward to detect, but subtle modifications are harder. In preliminary experiments, the margin separating fp8 from bf16 generation was small. The authors did not test whether TOPLOC distinguishes types of KV-cache compression.
- Tolerance leaves covert bandwidth (theoretical argument, in TOPLOC; https://trustbutveri.fyi/implementations/toploc/#flaw-4) [6]. TOPLOC accepts approximate matches. A check of this kind can put an upper bound on the covert bandwidth available to an adversary, but it cannot close that bandwidth. The limit applies to all statistical verification schemes.
- Low-communication training reduces the bandwidth training needs (theoretical argument, in Bandwidth limits and compartmentalization; https://trustbutveri.fyi/mechanisms/bandwidth-limits-and-compartmentalization/#flaw-1) [12][16][17]. DiLoCo matched fully synchronous training on 8 workers while communicating 500 times less. Rahman writes that this family of methods theoretically allows large-scale training with less than 100 Mbps. Lucid includes these methods in its bounds, but notes that extreme activation compression, architectures with unusually small inter-layer widths, or modular paradigms could erode the margin.
- Operator control of pod routing collapses the bound (theoretical argument, in Bandwidth limits and compartmentalization; https://trustbutveri.fyi/mechanisms/bandwidth-limits-and-compartmentalization/#flaw-2) [12]. Lucid's analysis finds that if the operator can freely assign pods to routers, it could dedicate a whole cell of 100 or more pods to one pipeline stage. The bound then falls to about 90–220x uncompressed and as low as about 25x with compression. The proposed mitigation, auditor-controlled random assignment that is periodically re-randomized, has not been implemented.
- Undeclared local storage raises per-pod capacity (theoretical argument, in Bandwidth limits and compartmentalization; https://trustbutveri.fyi/mechanisms/bandwidth-limits-and-compartmentalization/#flaw-3) [12]. More memory or storage per pod helps an adversary. Lucid requires per-pod storage to be declared, capped and physically inspected.
- Training within one pod is not covered (open question, in Bandwidth limits and compartmentalization; https://trustbutveri.fyi/mechanisms/bandwidth-limits-and-compartmentalization/#flaw-4) [12]. Lucid's bounds concern pre-training models larger than the pods are sized for. Training models that fit in one pod, fine-tuning and reinforcement-learning post-training within one pod are outside the modelled threat.
- Parallel scale-up switches are hard enforcement points (theoretical argument, in Bandwidth limits and compartmentalization; https://trustbutveri.fyi/mechanisms/bandwidth-limits-and-compartmentalization/#flaw-5) [15]. In GB200 topologies, GPUs reach GPUs in other nodes through NVSwitches without a NIC on the path. Amodo notes that limits are hard to enforce there because many switches work in parallel, so compromising one or two would bypass the limit.

**Family finding context**

- Context for TOPLOC; applicability depends on the finding's scope. Tolerance for numerical noise leaves a covert channel (demonstrated attack, in Sampled inference recomputation; https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/#flaw-1) [6][7][8]. 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 TOPLOC; applicability depends on the finding's scope. Only recorded traffic is checked (theoretical argument, in Sampled inference recomputation; https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/#flaw-2) [7][9]. 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 TOPLOC; applicability depends on the finding's scope. Some inference optimizations are not covered (theoretical argument, in Sampled inference recomputation; https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/#flaw-3) [1][10]. 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 TOPLOC; applicability depends on the finding's scope. 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.

- **Tamper evidence for verifier devices** (R2 Demonstrated, assessed for detecting probing of proposed verifier hardware, using server and electronics prototypes as evidence)
  - Bandwidth limits and compartmentalization waits on it: Shaping devices and routing assignments must be trusted by both parties; Amodo has not yet fully analysed resilience to a compromised DPU.
- **Network taps and certifiers** (R1 Proposed, assessed for committing a complete record of cluster traffic, so declared inference can be checked)
  - Bandwidth limits and compartmentalization waits on it: The verifier must know that all traffic leaving a pod crosses the capped, monitored links.


## Dependencies

**Missing prerequisites**

- Tamper evidence for verifier devices (R2 Demonstrated, assessed for detecting probing of proposed verifier hardware, using server and electronics prototypes as evidence), needed by Bandwidth limits and compartmentalization

**Blockers**

- Sampled inference recomputation: No independent security evaluation has been published, and Amodo Design rates red-teaming of recomputation schemes as 'not started'. (adversarial validation) [18]
- Sampled inference recomputation: The verifier must run the model itself, which suits the paper's setting of providers serving open-weights models. (privacy & leakage) [1]
- Bandwidth limits and compartmentalization: No cap that a verifier can check has been implemented or red-teamed. (adversarial validation) [12]
- Bandwidth limits and compartmentalization: The verifier must know that all traffic leaving a pod crosses the capped, monitored links. (coverage & hidden compute; waits on Network taps and certifiers) [13]
- Bandwidth limits and compartmentalization: Shaping devices and routing assignments must be trusted by both parties; Amodo has not yet fully analysed resilience to a compromised DPU. (hardware trust; waits on Tamper evidence for verifier devices) [12][15]
- Bandwidth limits and compartmentalization: Advances in low-communication training could shrink the margin that the cap enforces. (capacity bounds) [12][16][17]


## What the verifier sees

- Model weights: shown by none; depends on the design for none; hidden by none; not involved in Bandwidth limits and compartmentalization; unspecified for Sampled inference recomputation.
- Inputs and outputs: shown by none; depends on the design for none; hidden by none; not involved in Bandwidth limits and compartmentalization; unspecified for Sampled inference recomputation.
- Training data: shown by none; depends on the design for none; hidden by none; not involved in Bandwidth limits and compartmentalization; unspecified for Sampled inference recomputation.

## Implementations

- 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)
- Bandwidth limits and compartmentalization: [AI 2040 inference-only verification stack](https://trustbutveri.fyi/implementations/ai-2040-inference-only-verification-plan/) (R1, proposed architecture); [RAND secure inference data center (SIDC) design](https://trustbutveri.fyi/implementations/rand-secure-inference-data-centers/) (R1, proposed architecture)

## Sources

1. TOPLOC: A Locality Sensitive Hashing Scheme for Trustless Verifiable Inference, J. M. Ong et al. (2025). https://proceedings.mlr.press/v267/ong25a.html
2. PrimeIntellect-ai/toploc (GitHub repository), Prime Intellect (2025). https://github.com/PrimeIntellect-ai/toploc
3. INTELLECT-2: A Reasoning Model Trained Through Globally Decentralized Reinforcement Learning, Prime Intellect Team et al. (2025). https://arxiv.org/abs/2505.07291
4. SYNTHETIC-2, Prime Intellect (2025). https://www.primeintellect.ai/blog/synthetic-2
5. SYNTHETIC-2 Release: Four Million Collaboratively Generated Reasoning Traces, Prime Intellect (2025). https://www.primeintellect.ai/blog/synthetic-2-release
6. Bit-Exact AI Inference Verification Without Performance Tradeoffs, N. Cankaya (2026). https://arxiv.org/abs/2606.00279
7. Verifying LLM Inference to Detect Model Weight Exfiltration, R. Rinberg et al. (2025). https://arxiv.org/abs/2511.02620
8. Adversarial Entropy Inflation Against Gumbel-Based Inference Verification, N. Kezins (2026). https://arxiv.org/abs/2608.23375
9. Example Schemes for Verifying High-Stakes AI Agreements, Amodo Design (2026). https://amododesign.com/notes/2026-06-23-verification-algorithms/
10. DiFR: Inference Verification Despite Nondeterminism, A. Karvonen et al. (2025). https://arxiv.org/abs/2511.20621
11. Verification Plan, R. Dean (2026). https://ai-2040.com/supplements/verification-plan
12. Traffic Shaping for Workload Classification, Lucid Computing (2026). https://lucidcomputing.substack.com/p/traffic-shaping-for-workload-classification
13. 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
14. De-risking Interconnect Limits for AI Verification, A. Scher et al. (2026). https://techgov.intelligence.org/blog/de-risking-interconnect-limits-for-ai-verification
15. The Tray as a Bandwidth Boundary, Amodo Design (2026). https://amododesign.com/notes/2026-03-16-dpu-bandwidth-limiter/
16. DiLoCo: Distributed Low-Communication Training of Language Models, A. Douillard et al. (2024). https://arxiv.org/abs/2311.08105
17. Does Distributed Training Undermine Compute Governance?, R. Rahman (2026). https://arxiv.org/abs/2605.29359
18. AI 2040 Plan A — Verification SITREP, Amodo Design (2026). https://amododesign.com/ai-verification/plan-a-sitrep/