Proposal Explorer

Add claims to see which ones the mechanisms address.

In production

• R3 Deterministic and bit-exact inference for reproducing open-model inference from receipts in Gensyn's information-market service

Built for an adversarial prover

Deterministic and bit-exact inference

No new hardware needed

Deterministic and bit-exact inference

Not counted

Excluded by your filters: Model identity attestation

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

• Model identity attestation, Attestable Audits: Analysis · excluded by filters
• Deterministic and bit-exact inference: Analysis

Excluded by your filters

• Model identity attestation Readiness R2 is below the minimum of R3.

Family findings

• Model identity attestation

  Context for Attestable Audits. These findings concern the family or other implementations; applicability must be checked against their stated scope.

  • Underlying attestation can be forged or relayed in Model identity attestation

    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. ^{3 4 5 6 7 8}

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

    Demonstrated attack · Critical · Open · Inherited finding. On the record · Related finding in TEE remote attestation for AI workloads

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

  • Launch-state attestation does not by itself cover weights loaded later in Model identity attestation

    Attestation measures launch state, and weights are read from disk after boot. A signature checked at load time does not stop a malicious hypervisor from altering the disk afterwards. Tinfoil reports mitigating this with dm-verity checks on every read. Unmeasured runtime configuration remains a general risk. ^{9 10}

    Theoretical argument · Significant · Mitigated. On the record

  • For private models, a user can confirm consistency but not content in Model identity attestation

    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. ^{1 9}

    Open question · Significant · Open. On the record

  • Recomputation depends on trusted logging and randomness, and its tolerance leaves a covert channel in Model identity attestation

    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. ^{11 12}

    Demonstrated attack · Significant · Open. On the record

    Related mechanism R1 Network taps and certifiers: 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. A pointer, not evidence that this flaw is mitigated. Add

    Related mechanism R3 Deterministic and bit-exact inference: Bit-exact inference would remove the numerical tolerance that leaves this channel. A pointer, not evidence that this flaw is mitigated. In the proposal.

Open significant flaws

3 flaws in 2 mechanisms

• Relies on the TEE vendor and inherits TEE attacks in Attestable Audits

  The prototype trusts AWS Nitro, not the Intel TDX or AMD SEV-SNP attestation roots targeted by the cited confidential-VM studies. Those studies are class context, not a demonstrated attack on this Nitro prototype.

  The design depends on trusting the TEE vendor, AWS in the prototype. The authors cite memory-aliasing, ciphertext side-channel and malicious-interrupt attacks on confidential VMs (BadRAM, CIPHERLEAKS, Heckler). Their answer is to revoke vulnerable base images once such attacks are discovered. ¹

  Response: The authors propose revoking vulnerable base images; they do not report a red-team evaluation of the prototype.

  Theoretical argument · Significant · Open · Inherited finding. On the record · Related finding in TEE remote attestation for AI workloads

• Prompt-based model exfiltration is a residual gap in Attestable Audits

  The authors state that "prompt-based model exfiltration during the user interaction step remains a residual gap". ¹

  Open question · Significant · Open. On the record

• Cross-hardware replay relies on reverse-engineered, closed behaviour in Deterministic and bit-exact inference

  Emulating one GPU's rounding on another requires reverse-engineering tensor-core arithmetic and modelling proprietary kernel choices. Hawkeye covers a subset of NVIDIA architectures and states that attention and other higher-level operations need further reverse engineering. For the bit-exact emulator, a proprietary Hopper kernel family is an open edge case. ^{13 15}

  Open question · Significant · Open. On the record

Open minor flaws

2 mechanisms with minor findings

• CPU-only enclaves force small, quantized models and high cost in Attestable Audits

  Memory limits required 4-bit quantization, and the quantized model scored 51.4% on zero-shot MMLU. CPU inference cost 21.7 times as much per token as GPU inference, and the enclave roughly doubled the CPU cost. The authors wrote that H100 confidential computing had no multi-GPU support. NVIDIA's white paper of August 2025 describes a protected-PCIe mode that passes all eight GPUs of a Hopper HGX node to one confidential VM, with NVLink traffic unencrypted. ^{1 2}

  Open question · Minor · Open. On the record

• Some kernels remain genuinely nondeterministic in Deterministic and bit-exact inference

  The bit-exact work separates kernels that are deterministic but not batch-invariant from truly nondeterministic ones that use atomic functions. Some integer de-quantization kernels use atomic additions and remain nondeterministic, so exact replay needs backends that avoid them. ¹³

  Open question · Minor · Open. On the record

Mechanisms on the map that are not in the proposal. Pointers, not recommendations. ?

• R3 TEE remote attestation for AI workloads Add
  • Model identity attestation waits on it. The prototype needs porting to GPU confidential computing to handle larger models; the authors expect an overhead as small as 5 times there.

Missing prerequisites

• R3 TEE remote attestation for AI workloads needed by Model identity attestation Add

Blockers

6 blockers recorded

• Model identity attestation
  • The prototype needs porting to GPU confidential computing to handle larger models; the authors expect an overhead as small as 5 times there. Performance & compatibility. Waits on TEE remote attestation for AI workloads ¹
  • As of September 2026 no code has been released for the prototype. Adversarial validation ¹
• Deterministic and bit-exact inference
  • Batch-invariant kernels cost throughput: in Thinking Machines' Qwen3-8B test, an improved deterministic build took 42 s against 26 s for vLLM's default, and SGLang reports an average 34.35% slowdown on its FlashInfer and FlashAttention 3 backends. Performance & compatibility ^{14 16}
  • Coverage is incomplete: the bit-exact emulator targets dense blocks on NVIDIA GPUs and excludes mixture-of-experts inference and training, and vLLM's batch-invariant mode is in beta, with open work on AMD hardware and speculative decoding. Performance & compatibility ^{13 17 24}
  • Amodo's status page for the AI 2040 verification plan rates a reproducible inference stack for that plan as 'not started'. Performance & compatibility ²⁵
  • Exact replay requires the prover to disclose weights, software versions, parallelism and batch sizes to whoever recomputes. Privacy & leakage ^{13 23}

From the family or selected implementation's record. ?

Model weights

Depends on the design for Deterministic and bit-exact inference.

Unspecified for Model identity attestation. Check the implementation record.

Inputs and outputs

Depends on the design for Deterministic and bit-exact inference.

Unspecified for Model identity attestation. Check the implementation record.

Training data

Not involved: Deterministic and bit-exact inference.

Unspecified for Model identity attestation. Check the implementation record.

Model identity attestation

• R2 Attestable Audits Research prototype, University of Cambridge
• R2 PAL*M Research prototype, University of Waterloo
• R3 Tinfoil model identity (Modelwrap) Product, Tinfoil

Deterministic and bit-exact inference

• R2 Batch-invariant inference kernels (Thinking Machines) Open-source project, Thinking Machines Lab
• R3 Verde and RepOps (Gensyn) Product, Gensyn
• R1 Low-trust AI compute verification system overview Proposed architecture, Machine Intelligence Research Institute

1. Attestable Audits: Verifiable AI Safety Benchmarks Using Trusted Execution Environments, C. Schnabl et al. (2025). Original
2. NVIDIA Secure AI with Blackwell and Hopper GPUs (White Paper), NVIDIA (2025). Original
3. TEE.fail: Breaking Trusted Execution Environments via DDR5 Memory Bus Interposition, J. Chuang et al. (2026). Original
4. Battering RAM: Low-Cost Interposer Attacks on Confidential Computing via Dynamic Memory Aliasing, J. De Meulemeester et al. (2026). Original
5. RMPocalypse: How a Catch-22 Breaks AMD SEV-SNP, B. Schlüter & S. Shinde (2025). Original
6. SEV-SNP RMP Initialization Vulnerability (AMD-SB-3020), AMD (2025). Original
7. A primer on secure enclaves, Tinfoil (2026). Original
8. PAL*M: Property Attestation for Large Generative Models, P. Chantasantitam et al. (2026). Original
9. How Tinfoil Proves Exactly What Model Is Running, Tinfoil Team (2026). Original
10. On TEEs for Privacy-Preserving Monitoring in AI Governance, Gloria Z (2026). Original
11. Verifying LLM Inference to Detect Model Weight Exfiltration, R. Rinberg et al. (2025). Original
12. Adversarial Entropy Inflation Against Gumbel-Based Inference Verification, N. Kezins (2026). Original
13. Bit-Exact AI Inference Verification Without Performance Tradeoffs, N. Cankaya (2026). Original
14. Defeating Nondeterminism in LLM Inference, H. He & Thinking Machines Lab (2025). Original
15. Hawkeye: Reproducing GPU-Level Non-Determinism, E. Badash et al. (2026). Original
16. Towards Deterministic Inference in SGLang and Reproducible RL Training, The SGLang Team (2025). Original
17. Batch Invariance (vLLM documentation), vLLM project (2026). Original
18. gensyn-ai/ree: Gensyn Reproducible Execution Environment (GitHub repository), Gensyn (2026). Original
19. EigenCloud Brings Verifiable AI to Mass Market with EigenAI and EigenCompute Launches, EigenCloud (2025). Original
20. Building Delphi: Pricing, Settlement, and Agentic Trading, D. Jedamski (2026). Original
21. Reproducible Execution Environment (REE) (Gensyn documentation), Gensyn (2026). Original
22. What is Delphi? (Delphi documentation), Gensyn (2026). Original
23. A System Overview for Near-Term, Low-Trust AI Compute Verification, N. Cankaya (2026). Original
24. [Feature]: Batch Invariant Feature and Performance Optimization (vLLM issue #27433), vLLM project contributors (2025). Original
25. AI 2040 Plan A — Verification SITREP, Amodo Design (2026). Original