# 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-0023,M-0012&coop=partial

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.

- **Prover cooperation: Partial at most.** "Partial at most" removes mechanisms that need the prover's active participation. "Not required" keeps only those that work without it. Required: the prover takes part, for example by logging requests, producing proofs or opening records. Partial: some access, such as installing a device. Not required: works from outside, such as satellite imagery.

6 of 25 mechanisms on the map pass these filters.

## 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 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Safeguard attestation (excluded by the filters) | R2 | Semi-trusted | Analysis | Existing features | 0 / 5 / 0 | depends | depends | not involved |
| Model identity attestation (excluded by the filters) | R3 | Semi-trusted | Independent red-team | Existing features | 1 / 2 / 0 | depends | depends | not involved |

## Claims

No claims chosen.

## Mechanisms

### Safeguard attestation

Hardware-signed evidence that an AI service ran its declared safeguards, such as a guardrail classifier or monitor, when producing a given response. ([Safeguard attestation](https://trustbutveri.fyi/mechanisms/safeguard-attestation/))

- Assessment: mechanism family.
- Readiness: R2 Demonstrated, assessed for attesting that a declared safeguard mediated a service's responses.
- Claims in this proposal: none of them.
- Threat model: semi-trusted prover. Hardware: existing features. Prover cooperation: required. Attack testing: analysis. Category: Cryptographic & computational.
- What the verifier sees: model weights depends; inputs and outputs depends; training data not involved. The enclave route signs hashes of the safeguard, request and response; a low-trust design has the verifier re-run and screen sampled requests itself.
- Filter conflict: Prover cooperation is required; the filter allows partial at most.

### Model identity attestation

Establishes that responses come from a specific, committed set of model weights, using enclave measurements or recomputation of sampled outputs. ([Model identity attestation](https://trustbutveri.fyi/mechanisms/model-identity-attestation/))

- Assessment: mechanism family.
- Readiness: R3 In production, assessed for showing users that a service runs the declared model weights.
- Claims in this proposal: none of them.
- Threat model: semi-trusted prover. Hardware: existing features. Prover cooperation: required. Attack testing: independent red-team. Category: Cryptographic & computational.
- What the verifier sees: model weights depends; inputs and outputs depends; training data not involved. The enclave route shows only hashes; the recomputation route gives the verifier the weights and the sampled requests and responses.
- Filter conflict: Prover cooperation is required; the filter allows partial at most.


## Properties

Not counted as properties, because the filters exclude them: Safeguard attestation and Model identity attestation.


## 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**

- Safeguard attestation: Analysis (excluded by filters)
- Model identity attestation: Independent red-team (excluded by filters)


## Limits

**Excluded by the filters**

- Safeguard attestation: Prover cooperation is required; the filter allows partial at most.
- Model identity attestation: Prover cooperation is required; the filter allows partial at most.

**Open critical flaws**

- Underlying attestation can be forged or relayed (demonstrated attack, in Model identity attestation; https://trustbutveri.fyi/mechanisms/model-identity-attestation/#flaw-1) [4][7][10][11][12][16]. Inherited finding. Critical for the enclave route against an operator with physical access to affected hardware, or control of an unpatched SEV-SNP hypervisor. It does not apply to the recomputation route. PAL*M excludes physical attacks, and Tinfoil acknowledges this boundary. The enclave route inherits the platform-specific TEE attestation failures. Intel TDX forgery and H100 relay were demonstrated with physical access and host control. Battering RAM defeated AMD SEV-SNP attestation on DDR4 servers; RMPocalypse did so from malicious host software on platforms without AMD's fixes. These demonstrate failures of the trust roots, not of each model-commitment protocol. Related finding: https://trustbutveri.fyi/mechanisms/tee-remote-attestation/#flaw-1.

  Response: The TEE.fail authors report that physical interposer attacks are outside Intel's and AMD's threat models. AMD reports fixes for RMPocalypse.

  Related mechanism: Hardware-enabled guarantees (flexHEG) and guarantee processors (R1, not in the proposal). A tamper-protected enclosure around the chip is the proposed answer when the party that holds the hardware may attack it physically.

**Open significant flaws**

- Attestation shows a safeguard ran, not that it is effective (theoretical argument, in Safeguard attestation; https://trustbutveri.fyi/mechanisms/safeguard-attestation/#flaw-1) [1]. Proof of guardrail ensures that the guardrail executed, but the guardrail can still err or be jailbroken. Because the guardrail must be open source, a malicious developer can attack it with jailbreaks while still presenting a valid proof. In the authors' evaluation, Llama Guard 3 reached an F1 score of 0.56 on the unsafe class of the ToxicChat dataset. The authors state that proof of guardrail should not be interpreted or advertised as proof of safety.
- Selective attestation leaves traffic uncovered (theoretical argument, in Safeguard attestation; https://trustbutveri.fyi/mechanisms/safeguard-attestation/#flaw-2) [1][4][9]. Attestations are issued per response. In the prototype, the agent offers them when it receives high-stakes questions, so nothing shows that unattested traffic went through the same path. PAL*M's authors note that a prover could cherry-pick favourable executions, and suggest verifier-published nonces or requesting only session-level proofs. A governance analysis notes that auditors also need assurance that all activity is accounted for, since a host could start a second confidential virtual machine that bypasses monitoring.

  Related mechanism: On-chip telemetry from timing, memory and performance counters (R2, not in the proposal). On-chip counters are a proposed route to evidence about everything a chip runs, including a second virtual machine that skips the safeguard.
- Measurements may omit behaviour-relevant configuration or runtime changes (theoretical argument, in Safeguard attestation; https://trustbutveri.fyi/mechanisms/safeguard-attestation/#flaw-3) [9]. Every component that influences inference behaviour must be covered by the launch measurement, including feature flags, environment variables and invocation arguments. A launch measurement also does not show that a program keeps running as measured if the kernel is later compromised.
- Components outside the attested boundary (theoretical argument, in Safeguard attestation; https://trustbutveri.fyi/mechanisms/safeguard-attestation/#flaw-4) [1][2]. In the proof-of-guardrail experiments, the guardrail model and the agent's backend model were both reached through external APIs, and the authors leave the decision to trust those APIs to the verifier. The measured wrapper must also have no vulnerability that lets the unmeasured agent bypass the guardrail, for example by executing arbitrary commands inside the enclave. The code's README states that the enclave does not currently restrict the agent's arbitrary command execution, which could be used to bypass guardrails.
- Memory-bus interposition extracts attestation keys and forges attestations (demonstrated attack, in Safeguard attestation; https://trustbutveri.fyi/mechanisms/safeguard-attestation/#flaw-5) [1][4][7][8][9][10][11][12][13]. Inherited finding. Applies to variants using the affected Intel or AMD trust roots. PAL*M excludes physical attacks. A TDX-backed safeguard claim against a physical host attacker would be defeated, but these studies do not demonstrate a break of the AWS Nitro proof-of-guardrail prototype or of verifier-side recomputation. The TEE findings cover DDR5 attacks on Intel TDX, the H100 relay demonstration, DDR4 attacks on AMD SEV-SNP, and software-only SEV-SNP forgery before AMD's fixes. These are inherited hardware limits; a governance analysis explains why physical access matters in a treaty setting. Related finding: https://trustbutveri.fyi/mechanisms/tee-remote-attestation/#flaw-1.

  Response: Intel and AMD place the physical attack class outside their threat models, according to the researchers. AMD reports firmware fixes for RMPocalypse.

  Related mechanism: Hardware-enabled guarantees (flexHEG) and guarantee processors (R1, not in the proposal). A tamper-protected enclosure around the chip is the proposed answer when the party that holds the hardware may attack it physically.
- For private models, a user can confirm consistency but not content (open question, in Model identity attestation; https://trustbutveri.fyi/mechanisms/model-identity-attestation/#flaw-3) [14][21]. When weights are not published, users can check that the same root hash is served each time, but not what the model is. Pairing the hash with an attested evaluation, as in Attestable Audits, is one proposed remedy.
- Recomputation depends on trusted logging and randomness, and its tolerance leaves a covert channel (demonstrated attack, in Model identity attestation; https://trustbutveri.fyi/mechanisms/model-identity-attestation/#flaw-4) [15][20]. The recomputation variant assumes that every input, output and seed is logged correctly, and that the attacker can neither predict nor manipulate which messages are sampled for verification. Legitimate nondeterminism concentrates at a few token positions, and slow leaks within the tolerated slack remain possible. An independent study showed that an adversary who controls the prompts roughly doubles the bits leaked per token, reducing the exfiltration slowdown from 146–254 times under benign prompts to 60–118 times. The attack targets the exfiltration bound, not the check that outputs match the declared model.

  Related mechanism: Network taps and certifiers (R1, not in the proposal). Taps are proposed to copy and hash traffic on the monitored links, reducing reliance on the prover's own log. This still depends on the monitored boundary and trusted capture.

  Related mechanism: Deterministic and bit-exact inference (R3, not in the proposal). Bit-exact inference would remove the numerical tolerance that leaves this channel.


## 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.

- **On-chip telemetry from timing, memory and performance counters** (R2 Demonstrated, assessed for workload evidence from GPU counters and timing, assuming authentic measurements)
  - Bears on the open significant flaw "Selective attestation leaves traffic uncovered" in Safeguard attestation. On-chip counters are a proposed route to evidence about everything a chip runs, including a second virtual machine that skips the safeguard.
- **Hardware-enabled guarantees (flexHEG) and guarantee processors** (R1 Proposed, assessed for checking and enforcing training-compute limits on chips, against adversaries up to states). Excluded by the filters: prover cooperation required
  - Bears on the open critical flaw "Underlying attestation can be forged or relayed" in Model identity attestation. A tamper-protected enclosure around the chip is the proposed answer when the party that holds the hardware may attack it physically.
  - Bears on the open significant flaw "Memory-bus interposition extracts attestation keys and forges attestations" in Safeguard attestation. A tamper-protected enclosure around the chip is the proposed answer when the party that holds the hardware may attack it physically.
- **Deterministic and bit-exact inference** (R3 In production, assessed for reproducing open-model inference from receipts in Gensyn's information-market service). Excluded by the filters: prover cooperation required
  - Bears on the open significant flaw "Recomputation depends on trusted logging and randomness, and its tolerance leaves a covert channel" in Model identity attestation. Bit-exact inference would remove the numerical tolerance that leaves this channel.
  - Model identity attestation waits on it: Numerical nondeterminism limits how tightly recomputation can pin down the model and sampling.
- **Network taps and certifiers** (R1 Proposed, assessed for committing a complete record of cluster traffic, so declared inference can be checked). Excluded by the filters: prover cooperation required
  - Bears on the open significant flaw "Recomputation depends on trusted logging and randomness, and its tolerance leaves a covert channel" in Model identity attestation. Taps are proposed to copy and hash traffic on the monitored links, reducing reliance on the prover's own log. This still depends on the monitored boundary and trusted capture.
- **TEE remote attestation for AI workloads** (R3 In production, assessed for showing which software ran to a party that distrusts the operator holding the hardware). Excluded by the filters: prover cooperation required
  - Safeguard attestation waits on it: Frontier model inference typically needs several GPUs, GPU confidential computing is less mature than CPU support, and CPU inference, which an enclave prototype had to use, ran about 100 times slower than GPU inference.
  - Safeguard attestation waits on it: Trust rests on a small number of hardware vendors, and a per-CPU Intel attestation key has been extracted by physical attack.
  - Model identity attestation waits on it: Attestation that resists physical attackers, for the enclave variant.


## Dependencies

**Missing prerequisites**

- TEE remote attestation for AI workloads (R3 In production, assessed for showing which software ran to a party that distrusts the operator holding the hardware), needed by Safeguard attestation

**Shared foundations**

- TEE remote attestation for AI workloads, relied on by Safeguard attestation and Model identity attestation

**Blockers**

- Safeguard attestation: No published design shows that all of a provider's traffic passes through the attested safeguard path; current evidence covers individual attested responses. (coverage & hidden compute) [1][9]
- Safeguard attestation: Frontier model inference typically needs several GPUs, GPU confidential computing is less mature than CPU support, and CPU inference, which an enclave prototype had to use, ran about 100 times slower than GPU inference. (performance & compatibility; waits on TEE remote attestation for AI workloads) [9][21]
- Safeguard attestation: Trust rests on a small number of hardware vendors, and a per-CPU Intel attestation key has been extracted by physical attack. (hardware trust; waits on TEE remote attestation for AI workloads) [7][9]
- Safeguard attestation: Safeguard evidence must be bound to the model actually served, which depends on model-identity attestation. (evidence binding; waits on Model identity attestation) [14][21]
- Safeguard attestation: No independent red-team or audit of a safeguard-attestation system has been published, and the available prototypes are described by their authors as proofs of concept that have not been stress-tested by a counterparty. (adversarial validation) [2][6]
- Model identity attestation: Attestation that resists physical attackers, for the enclave variant. (hardware trust; waits on TEE remote attestation for AI workloads) [7]
- Model identity attestation: Numerical nondeterminism limits how tightly recomputation can pin down the model and sampling. (protocol soundness; waits on Deterministic and bit-exact inference) [15]
- Model identity attestation: The recomputation variant needs the verifier to hold the declared weights. (access & governance) [15]


## What the verifier sees

- Model weights: shown by none; depends on the design for Safeguard attestation and Model identity attestation; hidden by none; not involved in none; unspecified for none.
- Inputs and outputs: shown by none; depends on the design for Safeguard attestation and Model identity attestation; 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 Safeguard attestation and Model identity attestation; unspecified for none.

## Implementations

- Safeguard attestation: none on the map
- Model identity attestation: [Attestable Audits](https://trustbutveri.fyi/implementations/attestable-audits/) (R2, research prototype); [PAL*M](https://trustbutveri.fyi/implementations/palm/) (R2, research prototype); [Tinfoil model identity (Modelwrap)](https://trustbutveri.fyi/implementations/tinfoil-model-identity/) (R3, product)

## Sources

1. Proof-of-Guardrail in AI Agents and What (Not) to Trust from It, X. Jin et al. (2026). https://arxiv.org/abs/2603.05786
2. Verifiable-ClawGuard: proof-of-guardrail reference code, SaharaLabsAI (2026). https://github.com/SaharaLabsAI/Verifiable-ClawGuard
3. Safety Without Compromising on Privacy, D. McCann-Sayles et al. (2026). https://tinfoil.sh/blog/2026-09-14-safety-without-compromising-privacy
4. PAL*M: Property Attestation for Large Generative Models, P. Chantasantitam et al. (2026). https://arxiv.org/abs/2601.16199
5. Enabling Verifiably-Scoped Monitoring through Large Language Models and Trusted Compute, B. Penchas et al. (2026). https://icml.cc/virtual/2026/78630
6. Auditor-in-a-Box: Tools for Third-Party Auditing, R. Rinberg & B. Penchas (2026). https://www.lesswrong.com/posts/uWYk7MM9hAf9GEbGe/auditor-in-a-box-tools-for-third-party-auditing
7. TEE.fail: Breaking Trusted Execution Environments via DDR5 Memory Bus Interposition, J. Chuang et al. (2026). https://tee.fail/
8. DDRop: Active Memory Interposer Attacks on Confidential VMs by Dropping DDR5 Writes, J. De Meulemeester et al. (2026). https://ddropattack.eu/
9. On TEEs for Privacy-Preserving Monitoring in AI Governance, Gloria Z (2026). https://techgov.intelligence.org/blog/on-tees-for-privacy-preserving-monitoring-in-ai-governance
10. Battering RAM: Low-Cost Interposer Attacks on Confidential Computing via Dynamic Memory Aliasing, J. De Meulemeester et al. (2026). https://batteringram.eu/
11. RMPocalypse: How a Catch-22 Breaks AMD SEV-SNP, B. Schlüter & S. Shinde (2025). https://rmpocalypse.github.io/
12. SEV-SNP RMP Initialization Vulnerability (AMD-SB-3020), AMD (2025). https://www.amd.com/en/resources/product-security/bulletin/amd-sb-3020.html
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. How Tinfoil Proves Exactly What Model Is Running, Tinfoil Team (2026). https://tinfoil.sh/blog/2026-02-03-proving-model-identity
15. Verifying LLM Inference to Detect Model Weight Exfiltration, R. Rinberg et al. (2025). https://arxiv.org/abs/2511.02620
16. A primer on secure enclaves, Tinfoil (2026). https://docs.tinfoil.sh/verification/secure-enclave-primer
17. Backend infrastructure, Tinfoil (2026). https://docs.tinfoil.sh/verification/attestation-architecture
18. How verification works in Tinfoil, Tinfoil (2026). https://docs.tinfoil.sh/verification/verification-in-tinfoil
19. modelwrap: Reproducible dm-verity read-only image of Huggingface models, Tinfoil (2026). https://github.com/tinfoilsh/modelwrap
20. Adversarial Entropy Inflation Against Gumbel-Based Inference Verification, N. Kezins (2026). https://arxiv.org/abs/2608.23375
21. Attestable Audits: Verifiable AI Safety Benchmarks Using Trusted Execution Environments, C. Schnabl et al. (2025). https://arxiv.org/abs/2506.23706