{
  "schema_version": "1.3",
  "url": "https://trustbutveri.fyi/explorer/?mechanisms=M-0014,M-0020&implementations=M-0014:I-0011",
  "data_generated": "2026-10-09",
  "definitions": {
    "methodology": "https://trustbutveri.fyi/about/methodology/",
    "readiness": "https://trustbutveri.fyi/about/readiness/",
    "filters": [
      {
        "id": "prover",
        "label": "Prover",
        "question": "How far can the party being checked be trusted?",
        "options": [
          {
            "value": "cooperative",
            "label": "Cooperative"
          },
          {
            "value": "semi-trusted",
            "label": "Semi-trusted"
          },
          {
            "value": "adversarial",
            "label": "Adversarial"
          }
        ],
        "rule": "Keeps mechanisms whose threat model holds against at least this prover. Adversarial is the strongest assumption.",
        "about": "The prover is the party being checked. Semi-trusted designs rely on part of its stack: usually the chip vendor's hardware root of trust, its firmware or counters, or its supply-chain records. Adversarial designs aim to hold even if it cheats wherever the checks allow, within their stated assumptions."
      },
      {
        "id": "onsite",
        "label": "Verifier devices on site",
        "question": "May the verifier install its own hardware at the prover's sites?",
        "options": [
          {
            "value": "no",
            "label": "Not allowed"
          }
        ],
        "rule": "\"Not allowed\" removes mechanisms that need a retrofit device, such as a network tap or a sealed sensor.",
        "about": "Some mechanisms need a device the verifier owns or trusts at the prover's facility, such as a network tap, a bandwidth limiter or a sealed sensor. Choose Not allowed when the setting rules that out. Inspectors are not covered."
      },
      {
        "id": "coop",
        "label": "Prover cooperation",
        "question": "How much must the prover take part?",
        "options": [
          {
            "value": "partial",
            "label": "Partial at most"
          },
          {
            "value": "none",
            "label": "Not required"
          }
        ],
        "rule": "\"Partial at most\" removes mechanisms that need the prover's active participation. \"Not required\" keeps only those that work without it.",
        "about": "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."
      },
      {
        "id": "chips",
        "label": "Chips",
        "question": "May the proposal depend on new chip designs?",
        "options": [
          {
            "value": "existing",
            "label": "Existing chips only"
          }
        ],
        "rule": "\"Existing chips only\" removes mechanisms that need changes to future chip designs.",
        "about": "New chip features take years to reach a deployed fleet and cover only chips made after they ship. Mechanisms that use shipping features, such as trusted execution environments or performance counters, stay."
      },
      {
        "id": "ready",
        "label": "Minimum development status",
        "question": "Development status",
        "options": [
          {
            "value": "R1",
            "label": "Proposed"
          },
          {
            "value": "R2",
            "label": "Research demonstration"
          },
          {
            "value": "R3",
            "label": "Operational use"
          },
          {
            "value": "R4",
            "label": "Legacy independent-evaluation filter",
            "legacy": true
          }
        ],
        "rule": "Keeps mechanisms whose readiness level is at least this one.",
        "about": "A level describes the public evidence for a mechanism's stated use, not its cost or feasibility. R3 can still have open critical flaws."
      },
      {
        "id": "tested",
        "label": "Attack testing",
        "question": "How hard has each mechanism been attacked in public?",
        "options": [
          {
            "value": "analysis",
            "label": "Published analysis"
          },
          {
            "value": "red-teamed",
            "label": "Red-teamed"
          },
          {
            "value": "independent-red-team",
            "label": "Independent red-team"
          }
        ],
        "rule": "Keeps mechanisms whose strongest published attack testing is at least this.",
        "about": "The strongest published attempt to break the mechanism for its verification use: a security analysis, red-teaming by its developers or collaborators, or a red team independent of them."
      },
      {
        "id": "hide",
        "label": "Keep hidden from the verifier",
        "question": "What must the verifier never see?",
        "options": [
          {
            "value": "weights",
            "label": "Model weights"
          },
          {
            "value": "io",
            "label": "Inputs and outputs"
          },
          {
            "value": "training",
            "label": "Training data"
          }
        ],
        "rule": "Removes mechanisms that show the asset to the verifier. Conditional or unspecified exposure stays with a note and needs checking against the privacy requirement.",
        "about": "Model weights: the checked model's parameters. Inputs and outputs: the requests a deployed model serves and its responses. Training data: what a model was trained on. Each mechanism's exposure is the editors' reading of its record: shown, depends on the design (kept, with a note), hidden, not involved, or unspecified for a selected implementation. Code and configuration are not covered yet."
      }
    ],
    "exposure": "For model weights, inputs and outputs, and training data. This is the editors' reading of each mechanism's record (its threat model, how it works and its limitations), not a field of the record. Shown: the verifier sees it. Depends: on the design or variant, or the verifier sees only samples. Hidden: the verifier sees only commitments, hashes, proofs or results. Not involved: the record does not handle it. Unspecified: the selected implementation has no asset-specific assessment here.",
    "claim_status": {
      "addressed": "A mechanism in the proposal is aimed at this claim and is not excluded by the filters.",
      "partly-addressed": "Only supporting mechanisms, or mechanisms aimed at it that the filters exclude.",
      "unaddressed": "No mechanism in the proposal addresses this claim."
    },
    "finding_classification": {
      "failure": {
        "label": "Known failures",
        "singular": "Known failure",
        "anchor": "known-flaws"
      },
      "scope-limitation": {
        "label": "Scope limitations",
        "singular": "Scope limitation",
        "anchor": "scope-limitations"
      },
      "open-question": {
        "label": "Open questions",
        "singular": "Open question",
        "anchor": "open-questions"
      }
    },
    "finding_scope": "Evidence scope describes where a finding was demonstrated; it does not establish applicability to every implementation in the mechanism family.",
    "claim_finding_scope": "open_critical_findings names active failures on the assessed records; open_critical_context names conditional family failures whose implementation applicability is unassessed.",
    "legacy_status": "The status field retains covered/partial/none for compatibility. It names claim links, never successful verification. Use claim_status and status_label for presentation."
  },
  "filters": {
    "prover": "",
    "onsite": "",
    "coop": "",
    "chips": "",
    "ready": "",
    "tested": "",
    "hide": []
  },
  "mechanisms_passing_filters": 25,
  "claims": [],
  "mechanisms": [
    {
      "id": "M-0014",
      "title": "Bandwidth limits and compartmentalization",
      "url": "https://trustbutveri.fyi/mechanisms/bandwidth-limits-and-compartmentalization/",
      "assessment_record": {
        "id": "I-0011",
        "title": "AI 2040 inference-only verification stack",
        "url": "https://trustbutveri.fyi/implementations/ai-2040-inference-only-verification-plan/"
      },
      "finding_counts": {
        "failure": 1,
        "scope_limitation": 2,
        "open_question": 0,
        "open_failures": {
          "critical": 0,
          "significant": 1,
          "minor": 0
        }
      },
      "selected_implementation": {
        "id": "I-0011",
        "title": "AI 2040 inference-only verification stack",
        "url": "https://trustbutveri.fyi/implementations/ai-2040-inference-only-verification-plan/"
      },
      "readiness": {
        "level": "R1",
        "scope": "showing that retrofitted data centres run only inference",
        "confidence": "medium",
        "evidence": [
          "S-0067",
          "S-1511",
          "S-1512",
          "S-1008",
          "S-1312"
        ]
      },
      "development_status": {
        "code": "R1",
        "label": "Proposed",
        "short": "Proposed",
        "rank": 1,
        "legacy_code": "R1"
      },
      "security_evidence": {
        "attack_testing": {
          "status": "none",
          "label": "No published adversarial analysis recorded",
          "kind": "none"
        },
        "independent_evaluation": {
          "status": "unassessed"
        },
        "formal_proof": {
          "status": "unassessed"
        },
        "deployment_assurance": {
          "status": "unassessed"
        },
        "legacy_evaluation_code": null,
        "scoped_findings": [
          {
            "n": 3,
            "severity": "significant",
            "status": "open",
            "evidence_scope": "unassessed",
            "sources": [
              "S-1511",
              "S-0015",
              "S-1507",
              "S-0067"
            ]
          }
        ],
        "open_failures": {
          "critical": 0,
          "significant": 1,
          "minor": 0
        }
      },
      "assessed_properties": {
        "threat_model": "adversarial",
        "hardware_requirement": "retrofit-device",
        "prover_cooperation": "required",
        "adversarial_evaluation": "none"
      },
      "claims": [],
      "exposure": {
        "weights": "unknown",
        "io": "unknown",
        "training": "unknown",
        "note": "This Explorer has no asset-specific exposure assessment for this implementation. Check its source and deployment assumptions.",
        "sources": []
      },
      "family_finding_context": [
        {
          "n": 1,
          "historical": false,
          "title": "Low-communication training reduces the bandwidth training needs",
          "classification": "open-question",
          "kind": "theoretical-argument",
          "severity": "significant",
          "status": "open",
          "evidence_scope": null,
          "scope_note": null,
          "related_finding": null,
          "description": "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.",
          "response": null,
          "sources": [
            "S-1314",
            "S-0060",
            "S-1301"
          ],
          "record": "M-0014",
          "represented_by": []
        },
        {
          "n": 2,
          "historical": false,
          "title": "Operator control of pod routing collapses the bound",
          "classification": "failure",
          "kind": "theoretical-argument",
          "severity": "significant",
          "status": "open",
          "evidence_scope": null,
          "scope_note": null,
          "related_finding": null,
          "description": "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.",
          "response": null,
          "sources": [
            "S-1301"
          ],
          "record": "M-0014",
          "represented_by": []
        },
        {
          "n": 3,
          "historical": false,
          "title": "Undeclared local storage raises per-pod capacity",
          "classification": "scope-limitation",
          "kind": "theoretical-argument",
          "severity": "significant",
          "status": "open",
          "evidence_scope": null,
          "scope_note": null,
          "related_finding": null,
          "description": "More memory or storage per pod helps an adversary. Lucid requires per-pod storage to be declared, capped and physically inspected.",
          "response": null,
          "sources": [
            "S-1301"
          ],
          "record": "M-0014",
          "represented_by": []
        },
        {
          "n": 4,
          "historical": false,
          "title": "Training within one pod is not covered",
          "classification": "scope-limitation",
          "kind": "open-question",
          "severity": "significant",
          "status": "open",
          "evidence_scope": null,
          "scope_note": null,
          "related_finding": null,
          "description": "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.",
          "response": null,
          "sources": [
            "S-1301"
          ],
          "record": "M-0014",
          "represented_by": []
        },
        {
          "n": 5,
          "historical": false,
          "title": "Parallel scale-up switches are hard enforcement points",
          "classification": "failure",
          "kind": "theoretical-argument",
          "severity": "significant",
          "status": "open",
          "evidence_scope": null,
          "scope_note": null,
          "related_finding": null,
          "description": "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.",
          "response": null,
          "sources": [
            "S-1313"
          ],
          "record": "M-0014",
          "represented_by": []
        }
      ],
      "filter_issues": []
    },
    {
      "id": "M-0020",
      "title": "Remote detection of data centres",
      "url": "https://trustbutveri.fyi/mechanisms/remote-detection-of-data-centres/",
      "assessment_record": {
        "id": "M-0020",
        "title": "Remote detection of data centres",
        "url": "https://trustbutveri.fyi/mechanisms/remote-detection-of-data-centres/"
      },
      "finding_counts": {
        "failure": 0,
        "scope_limitation": 2,
        "open_question": 1,
        "open_failures": {
          "critical": 0,
          "significant": 0,
          "minor": 0
        }
      },
      "selected_implementation": null,
      "readiness": {
        "level": "R1",
        "scope": "finding undeclared data centres above an agreed compute threshold",
        "confidence": "medium",
        "evidence": [
          "S-1410",
          "S-0002",
          "S-1409",
          "S-1411",
          "S-3301"
        ]
      },
      "development_status": {
        "code": "R1",
        "label": "Proposed",
        "short": "Proposed",
        "rank": 1,
        "legacy_code": "R1"
      },
      "security_evidence": {
        "attack_testing": {
          "status": "analysis",
          "label": "Published security analysis",
          "kind": "analysis"
        },
        "independent_evaluation": {
          "status": "unassessed"
        },
        "formal_proof": {
          "status": "unassessed"
        },
        "deployment_assurance": {
          "status": "unassessed"
        },
        "legacy_evaluation_code": null,
        "scoped_findings": [],
        "open_failures": {
          "critical": 0,
          "significant": 0,
          "minor": 0
        }
      },
      "assessed_properties": {
        "threat_model": "adversarial",
        "hardware_requirement": "none",
        "prover_cooperation": "not-required",
        "adversarial_evaluation": "analysis"
      },
      "claims": [],
      "exposure": {
        "weights": "none",
        "io": "none",
        "training": "none",
        "note": "Works from outside the facility; it does not handle model data."
      },
      "family_finding_context": [],
      "filter_issues": []
    }
  ],
  "strengths": {
    "covered": [],
    "production": [],
    "operationalUse": [],
    "adversarial": [
      "M-0014",
      "M-0020"
    ],
    "noNewHardware": [
      "M-0020"
    ],
    "mitigated": [],
    "notCounted": []
  },
  "properties": {
    "covered": [],
    "production": [],
    "operationalUse": [],
    "adversarial": [
      "M-0014",
      "M-0020"
    ],
    "noNewHardware": [
      "M-0020"
    ],
    "mitigated": [],
    "notCounted": []
  },
  "attack_testing": [
    {
      "id": "M-0020",
      "record": "M-0020",
      "evaluation": "analysis",
      "in_setting": true
    }
  ],
  "selected_implementations": {
    "M-0014": "I-0011"
  },
  "weaknesses": {
    "gaps": [],
    "excluded": [],
    "unlinked": [],
    "critical": [],
    "significant": [
      {
        "mech": "M-0014",
        "n": 3,
        "historical": false,
        "title": "A recomputation family degrades against prompt-controlling adversaries",
        "classification": "failure",
        "kind": "demonstrated-attack",
        "severity": "significant",
        "status": "open",
        "evidence_scope": null,
        "scope_note": null,
        "related_finding": null,
        "description": "The plan's companion page names DiFR among the recomputation schemes being tested. An independent study found that Gumbel-based inference verification, the family that includes Token-DiFR, leaks roughly twice as many bits per token when the adversary chooses prompts that disrupt the text's structure, across six models of 1 to 32 billion parameters. The slowdown it imposes on a hidden-signalling adversary falls from 146–254 times under benign prompts to 60–118 times. The attack weakens the bound on hidden information in outputs, which the plan relies on to keep undeclared results from leaving.",
        "response": null,
        "sources": [
          "S-1511",
          "S-0015",
          "S-1507",
          "S-0067"
        ],
        "record": "I-0011"
      }
    ],
    "criticalMechanisms": [],
    "significantMechanisms": [
      "M-0014"
    ],
    "familyContext": [
      {
        "id": "M-0014",
        "implementation": "I-0011",
        "flaws": [
          {
            "n": 1,
            "historical": false,
            "title": "Low-communication training reduces the bandwidth training needs",
            "classification": "open-question",
            "kind": "theoretical-argument",
            "severity": "significant",
            "status": "open",
            "evidence_scope": null,
            "scope_note": null,
            "related_finding": null,
            "description": "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.",
            "response": null,
            "sources": [
              "S-1314",
              "S-0060",
              "S-1301"
            ],
            "record": "M-0014",
            "represented_by": []
          },
          {
            "n": 2,
            "historical": false,
            "title": "Operator control of pod routing collapses the bound",
            "classification": "failure",
            "kind": "theoretical-argument",
            "severity": "significant",
            "status": "open",
            "evidence_scope": null,
            "scope_note": null,
            "related_finding": null,
            "description": "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.",
            "response": null,
            "sources": [
              "S-1301"
            ],
            "record": "M-0014",
            "represented_by": []
          },
          {
            "n": 3,
            "historical": false,
            "title": "Undeclared local storage raises per-pod capacity",
            "classification": "scope-limitation",
            "kind": "theoretical-argument",
            "severity": "significant",
            "status": "open",
            "evidence_scope": null,
            "scope_note": null,
            "related_finding": null,
            "description": "More memory or storage per pod helps an adversary. Lucid requires per-pod storage to be declared, capped and physically inspected.",
            "response": null,
            "sources": [
              "S-1301"
            ],
            "record": "M-0014",
            "represented_by": []
          },
          {
            "n": 4,
            "historical": false,
            "title": "Training within one pod is not covered",
            "classification": "scope-limitation",
            "kind": "open-question",
            "severity": "significant",
            "status": "open",
            "evidence_scope": null,
            "scope_note": null,
            "related_finding": null,
            "description": "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.",
            "response": null,
            "sources": [
              "S-1301"
            ],
            "record": "M-0014",
            "represented_by": []
          },
          {
            "n": 5,
            "historical": false,
            "title": "Parallel scale-up switches are hard enforcement points",
            "classification": "failure",
            "kind": "theoretical-argument",
            "severity": "significant",
            "status": "open",
            "evidence_scope": null,
            "scope_note": null,
            "related_finding": null,
            "description": "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.",
            "response": null,
            "sources": [
              "S-1313"
            ],
            "record": "M-0014",
            "represented_by": []
          }
        ]
      }
    ],
    "scopeLimitations": [
      {
        "mech": "M-0014",
        "n": 1,
        "historical": false,
        "title": "The recomputation server must be trusted",
        "classification": "scope-limitation",
        "kind": "theoretical-argument",
        "severity": "significant",
        "status": "open",
        "evidence_scope": null,
        "scope_note": null,
        "related_finding": null,
        "description": "The plan calls the integrity of the recomputation server an extremely important aspect, and its argument that sampling verifies all outputs assumes that the server's computations and outputs can be trusted. The companion page notes that the server sits inside the prover's facility, possibly under the prover's physical control, and that hardening it against integrity attacks needs significant research. Amodo rates recomputation-server security as not on track.",
        "response": null,
        "sources": [
          "S-0067",
          "S-1511",
          "S-1008"
        ],
        "record": "I-0011"
      },
      {
        "mech": "M-0014",
        "n": 2,
        "historical": false,
        "title": "Spare compute is not verified",
        "classification": "scope-limitation",
        "kind": "theoretical-argument",
        "severity": "significant",
        "status": "open",
        "evidence_scope": null,
        "scope_note": null,
        "related_finding": null,
        "description": "The plan states that it does not verify that spare compute is unused for unapproved workloads, because this seems very challenging. It relies instead on side-channel bounds and memory wipes, so that the only results that persist are verified inference outputs.",
        "response": null,
        "sources": [
          "S-0067"
        ],
        "record": "I-0011"
      },
      {
        "mech": "M-0020",
        "n": 1,
        "historical": false,
        "title": "Facilities can be disguised or hidden",
        "classification": "scope-limitation",
        "kind": "theoretical-argument",
        "severity": "significant",
        "status": "open",
        "evidence_scope": null,
        "scope_note": null,
        "related_finding": null,
        "description": "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.",
        "response": null,
        "sources": [
          "S-1410"
        ]
      },
      {
        "mech": "M-0020",
        "n": 2,
        "historical": false,
        "title": "Small sites may not be detectable",
        "classification": "scope-limitation",
        "kind": "theoretical-argument",
        "severity": "significant",
        "status": "open",
        "evidence_scope": null,
        "scope_note": null,
        "related_finding": null,
        "description": "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.",
        "response": null,
        "sources": [
          "S-1410",
          "S-1409"
        ],
        "helps": [
          {
            "by": "M-0019",
            "how": "Accounts for chips from the fab onwards, which does not depend on a site being visible."
          }
        ]
      }
    ],
    "openQuestions": [
      {
        "mech": "M-0020",
        "n": 3,
        "historical": false,
        "title": "Search for unknown sites is undemonstrated",
        "classification": "open-question",
        "kind": "open-question",
        "severity": "significant",
        "status": "open",
        "evidence_scope": null,
        "scope_note": null,
        "related_finding": null,
        "description": "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\".",
        "response": null,
        "sources": [
          "S-1409"
        ]
      }
    ],
    "minor": 0,
    "minorFindings": [],
    "minorBy": [],
    "notDemonstrated": [
      "M-0014",
      "M-0020"
    ],
    "newChip": []
  },
  "findings": [
    {
      "mech": "M-0014",
      "record": "I-0011",
      "n": 1,
      "historical": false,
      "title": "The recomputation server must be trusted",
      "classification": "scope-limitation",
      "kind": "theoretical-argument",
      "severity": "significant",
      "status": "open",
      "evidence_scope": null,
      "scope_note": null,
      "related_finding": null,
      "description": "The plan calls the integrity of the recomputation server an extremely important aspect, and its argument that sampling verifies all outputs assumes that the server's computations and outputs can be trusted. The companion page notes that the server sits inside the prover's facility, possibly under the prover's physical control, and that hardening it against integrity attacks needs significant research. Amodo rates recomputation-server security as not on track.",
      "response": null,
      "sources": [
        "S-0067",
        "S-1511",
        "S-1008"
      ]
    },
    {
      "mech": "M-0014",
      "record": "I-0011",
      "n": 2,
      "historical": false,
      "title": "Spare compute is not verified",
      "classification": "scope-limitation",
      "kind": "theoretical-argument",
      "severity": "significant",
      "status": "open",
      "evidence_scope": null,
      "scope_note": null,
      "related_finding": null,
      "description": "The plan states that it does not verify that spare compute is unused for unapproved workloads, because this seems very challenging. It relies instead on side-channel bounds and memory wipes, so that the only results that persist are verified inference outputs.",
      "response": null,
      "sources": [
        "S-0067"
      ]
    },
    {
      "mech": "M-0014",
      "record": "I-0011",
      "n": 3,
      "historical": false,
      "title": "A recomputation family degrades against prompt-controlling adversaries",
      "classification": "failure",
      "kind": "demonstrated-attack",
      "severity": "significant",
      "status": "open",
      "evidence_scope": null,
      "scope_note": null,
      "related_finding": null,
      "description": "The plan's companion page names DiFR among the recomputation schemes being tested. An independent study found that Gumbel-based inference verification, the family that includes Token-DiFR, leaks roughly twice as many bits per token when the adversary chooses prompts that disrupt the text's structure, across six models of 1 to 32 billion parameters. The slowdown it imposes on a hidden-signalling adversary falls from 146–254 times under benign prompts to 60–118 times. The attack weakens the bound on hidden information in outputs, which the plan relies on to keep undeclared results from leaving.",
      "response": null,
      "sources": [
        "S-1511",
        "S-0015",
        "S-1507",
        "S-0067"
      ]
    },
    {
      "mech": "M-0020",
      "record": "M-0020",
      "n": 1,
      "historical": false,
      "title": "Facilities can be disguised or hidden",
      "classification": "scope-limitation",
      "kind": "theoretical-argument",
      "severity": "significant",
      "status": "open",
      "evidence_scope": null,
      "scope_note": null,
      "related_finding": null,
      "description": "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.",
      "response": null,
      "sources": [
        "S-1410"
      ]
    },
    {
      "mech": "M-0020",
      "record": "M-0020",
      "n": 2,
      "historical": false,
      "title": "Small sites may not be detectable",
      "classification": "scope-limitation",
      "kind": "theoretical-argument",
      "severity": "significant",
      "status": "open",
      "evidence_scope": null,
      "scope_note": null,
      "related_finding": null,
      "description": "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.",
      "response": null,
      "sources": [
        "S-1410",
        "S-1409"
      ],
      "helps": [
        {
          "by": "M-0019",
          "how": "Accounts for chips from the fab onwards, which does not depend on a site being visible."
        }
      ]
    },
    {
      "mech": "M-0020",
      "record": "M-0020",
      "n": 3,
      "historical": false,
      "title": "Search for unknown sites is undemonstrated",
      "classification": "open-question",
      "kind": "open-question",
      "severity": "significant",
      "status": "open",
      "evidence_scope": null,
      "scope_note": null,
      "related_finding": null,
      "description": "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\".",
      "response": null,
      "sources": [
        "S-1409"
      ]
    }
  ],
  "possible_additions": [
    {
      "id": "M-0017",
      "title": "Tamper evidence for verifier devices",
      "url": "https://trustbutveri.fyi/mechanisms/tamper-evidence-for-verifier-devices/",
      "readiness": "R2",
      "development_status": {
        "code": "R2",
        "label": "Research demonstration",
        "short": "Research demo",
        "rank": 2,
        "legacy_code": "R2"
      },
      "security_evidence": {
        "attack_testing": {
          "status": "analysis",
          "label": "Published security analysis",
          "kind": "analysis"
        },
        "independent_evaluation": {
          "status": "unassessed"
        },
        "formal_proof": {
          "status": "unassessed"
        },
        "deployment_assurance": {
          "status": "unassessed"
        },
        "legacy_evaluation_code": null,
        "scoped_findings": [
          {
            "n": 1,
            "severity": "significant",
            "status": "open",
            "evidence_scope": "mechanism",
            "sources": [
              "S-1317",
              "S-1318"
            ]
          },
          {
            "n": 3,
            "severity": "significant",
            "status": "open",
            "evidence_scope": "mechanism",
            "sources": [
              "S-1315",
              "S-0052",
              "S-3261"
            ]
          }
        ],
        "open_failures": {
          "critical": 0,
          "significant": 2,
          "minor": 0
        }
      },
      "fits_filters": true,
      "filter_issues": [],
      "reasons": [
        {
          "kind": "blocker",
          "mech": "M-0014",
          "text": "Checking that taps are correctly installed and stay in place at scale is not a solved problem, and hardening the recomputation server inside the prover's facility needs significant research."
        }
      ]
    },
    {
      "id": "M-0015",
      "title": "Memory wiping and proofs of secure erasure",
      "url": "https://trustbutveri.fyi/mechanisms/memory-wiping-and-secure-erasure/",
      "readiness": "R1",
      "development_status": {
        "code": "R1",
        "label": "Proposed",
        "short": "Proposed",
        "rank": 1,
        "legacy_code": "R1"
      },
      "security_evidence": {
        "attack_testing": {
          "status": "analysis",
          "label": "Published security analysis",
          "kind": "analysis"
        },
        "independent_evaluation": {
          "status": "unassessed"
        },
        "formal_proof": {
          "status": "unassessed"
        },
        "deployment_assurance": {
          "status": "unassessed"
        },
        "legacy_evaluation_code": null,
        "scoped_findings": [
          {
            "n": 2,
            "severity": "significant",
            "status": "open",
            "evidence_scope": "unassessed",
            "sources": [
              "S-0032",
              "S-0018"
            ]
          }
        ],
        "open_failures": {
          "critical": 0,
          "significant": 1,
          "minor": 0
        }
      },
      "fits_filters": true,
      "filter_issues": [],
      "reasons": [
        {
          "kind": "blocker",
          "mech": "M-0014",
          "text": "Memory wiping may use existing algorithms, but hardware testing is at an early stage."
        }
      ]
    },
    {
      "id": "M-0013",
      "title": "Network taps and certifiers",
      "url": "https://trustbutveri.fyi/mechanisms/network-taps-and-certifiers/",
      "readiness": "R1",
      "development_status": {
        "code": "R1",
        "label": "Proposed",
        "short": "Proposed",
        "rank": 1,
        "legacy_code": "R1"
      },
      "security_evidence": {
        "attack_testing": {
          "status": "analysis",
          "label": "Published security analysis",
          "kind": "analysis"
        },
        "independent_evaluation": {
          "status": "unassessed"
        },
        "formal_proof": {
          "status": "unassessed"
        },
        "deployment_assurance": {
          "status": "unassessed"
        },
        "legacy_evaluation_code": null,
        "scoped_findings": [
          {
            "n": 1,
            "severity": "significant",
            "status": "open",
            "evidence_scope": "unassessed",
            "sources": [
              "S-1300",
              "S-1507"
            ]
          },
          {
            "n": 4,
            "severity": "significant",
            "status": "open",
            "evidence_scope": "unassessed",
            "sources": [
              "S-1312"
            ]
          },
          {
            "n": 6,
            "severity": "minor",
            "status": "mitigated",
            "evidence_scope": "unassessed",
            "sources": [
              "S-1300"
            ]
          }
        ],
        "open_failures": {
          "critical": 0,
          "significant": 2,
          "minor": 0
        }
      },
      "fits_filters": true,
      "filter_issues": [],
      "reasons": [
        {
          "kind": "blocker",
          "mech": "M-0014",
          "text": "Passive optical taps work at 400G, but the 800G and 1600G line rates now arriving in data centres are undemonstrated."
        }
      ]
    },
    {
      "id": "M-0022",
      "title": "Side-channel suppression for isolated facilities",
      "url": "https://trustbutveri.fyi/mechanisms/side-channel-suppression/",
      "readiness": "R1",
      "development_status": {
        "code": "R1",
        "label": "Proposed",
        "short": "Proposed",
        "rank": 1,
        "legacy_code": "R1"
      },
      "security_evidence": {
        "attack_testing": {
          "status": "analysis",
          "label": "Published security analysis",
          "kind": "analysis"
        },
        "independent_evaluation": {
          "status": "unassessed"
        },
        "formal_proof": {
          "status": "unassessed"
        },
        "deployment_assurance": {
          "status": "unassessed"
        },
        "legacy_evaluation_code": null,
        "scoped_findings": [],
        "open_failures": {
          "critical": 0,
          "significant": 0,
          "minor": 0
        }
      },
      "fits_filters": true,
      "filter_issues": [],
      "reasons": [
        {
          "kind": "blocker",
          "mech": "M-0014",
          "text": "There is no plan yet for quickly scaling side-channel defences on a frontier cluster; only early theoretical pieces exist."
        }
      ]
    },
    {
      "id": "M-0003",
      "title": "Whole-workload recomputation (reproducible packets)",
      "url": "https://trustbutveri.fyi/mechanisms/reproducible-computation-packets/",
      "readiness": "R1",
      "development_status": {
        "code": "R1",
        "label": "Proposed",
        "short": "Proposed",
        "rank": 1,
        "legacy_code": "R1"
      },
      "security_evidence": {
        "attack_testing": {
          "status": "none",
          "label": "No published adversarial analysis recorded",
          "kind": "none"
        },
        "independent_evaluation": {
          "status": "unassessed"
        },
        "formal_proof": {
          "status": "unassessed"
        },
        "deployment_assurance": {
          "status": "unassessed"
        },
        "legacy_evaluation_code": null,
        "scoped_findings": [],
        "open_failures": {
          "critical": 0,
          "significant": 0,
          "minor": 0
        }
      },
      "fits_filters": true,
      "filter_issues": [],
      "reasons": [
        {
          "kind": "blocker",
          "mech": "M-0014",
          "text": "A fully reproducible inference stack needs substantial software and tooling, and per-packet network reproducibility may need considerable software, firmware and possibly hardware work."
        }
      ]
    },
    {
      "id": "M-0001",
      "title": "Sampled inference recomputation",
      "url": "https://trustbutveri.fyi/mechanisms/sampled-inference-recomputation/",
      "readiness": "R3",
      "development_status": {
        "code": "R3",
        "label": "Operational use",
        "short": "Operational use",
        "rank": 3,
        "legacy_code": "R3"
      },
      "security_evidence": {
        "attack_testing": {
          "status": "analysis",
          "label": "Published security analysis",
          "kind": "analysis"
        },
        "independent_evaluation": {
          "status": "unassessed"
        },
        "formal_proof": {
          "status": "unassessed"
        },
        "deployment_assurance": {
          "status": "unassessed"
        },
        "legacy_evaluation_code": null,
        "scoped_findings": [
          {
            "n": 1,
            "severity": "significant",
            "status": "open",
            "evidence_scope": "unassessed",
            "sources": [
              "S-0020",
              "S-0015",
              "S-1507"
            ]
          },
          {
            "n": 3,
            "severity": "significant",
            "status": "open",
            "evidence_scope": "unassessed",
            "sources": [
              "S-1000",
              "S-0016"
            ]
          },
          {
            "n": 4,
            "severity": "minor",
            "status": "open",
            "evidence_scope": "unassessed",
            "sources": [
              "S-0016"
            ]
          }
        ],
        "open_failures": {
          "critical": 0,
          "significant": 2,
          "minor": 1
        }
      },
      "fits_filters": true,
      "filter_issues": [],
      "reasons": [
        {
          "kind": "prerequisite",
          "mech": "M-0014"
        }
      ]
    }
  ],
  "goal": null,
  "design": null,
  "dependencies": {
    "prerequisites": [
      {
        "id": "M-0001",
        "neededBy": [
          "M-0014"
        ]
      },
      {
        "id": "M-0003",
        "neededBy": [
          "M-0014"
        ]
      }
    ],
    "shared": [],
    "blockers": [
      {
        "mech": "M-0014",
        "n": 1,
        "historical": false,
        "text": "A fully reproducible inference stack needs substantial software and tooling, and per-packet network reproducibility may need considerable software, firmware and possibly hardware work.",
        "theme": "performance-compatibility",
        "blocked_by": "M-0003",
        "sources": [
          "S-1511"
        ],
        "inProposal": false
      },
      {
        "mech": "M-0014",
        "n": 2,
        "historical": false,
        "text": "Passive optical taps work at 400G, but the 800G and 1600G line rates now arriving in data centres are undemonstrated.",
        "theme": "performance-compatibility",
        "blocked_by": "M-0013",
        "sources": [
          "S-1511"
        ],
        "inProposal": false
      },
      {
        "mech": "M-0014",
        "n": 3,
        "historical": false,
        "text": "Checking that taps are correctly installed and stay in place at scale is not a solved problem, and hardening the recomputation server inside the prover's facility needs significant research.",
        "theme": "hardware-trust",
        "blocked_by": "M-0017",
        "sources": [
          "S-1511",
          "S-1008"
        ],
        "inProposal": false
      },
      {
        "mech": "M-0014",
        "n": 4,
        "historical": false,
        "text": "There is no plan yet for quickly scaling side-channel defences on a frontier cluster; only early theoretical pieces exist.",
        "theme": "coverage-hidden-compute",
        "blocked_by": "M-0022",
        "sources": [
          "S-1511"
        ],
        "inProposal": false
      },
      {
        "mech": "M-0014",
        "n": 5,
        "historical": false,
        "text": "Memory wiping may use existing algorithms, but hardware testing is at an early stage.",
        "theme": "coverage-hidden-compute",
        "blocked_by": "M-0015",
        "sources": [
          "S-1511"
        ],
        "inProposal": false
      },
      {
        "mech": "M-0014",
        "n": 6,
        "historical": false,
        "text": "Robust red-teaming of recomputation schemes has not started, and most algorithm development remains academic.",
        "theme": "adversarial-validation",
        "blocked_by": null,
        "sources": [
          "S-1511",
          "S-1008"
        ],
        "inProposal": null
      },
      {
        "mech": "M-0020",
        "n": 1,
        "historical": false,
        "text": "Wide-area, automated detection of data centres is not yet practical and needs large training datasets.",
        "theme": "coverage-hidden-compute",
        "blocked_by": null,
        "sources": [
          "S-1409"
        ],
        "inProposal": null
      },
      {
        "mech": "M-0020",
        "n": 2,
        "historical": false,
        "text": "No measured detection or false-alarm rates for finding undeclared facilities have been published.",
        "theme": "adversarial-validation",
        "blocked_by": null,
        "sources": [
          "S-1410",
          "S-1409"
        ],
        "inProposal": null
      },
      {
        "mech": "M-0020",
        "n": 3,
        "historical": false,
        "text": "Recent high-resolution imagery is costly, is limited by weather and needs trained analysts.",
        "theme": "access-governance",
        "blocked_by": null,
        "sources": [
          "S-1409"
        ],
        "inProposal": null
      }
    ]
  },
  "exposure": {
    "weights": {
      "shown": [],
      "partial": [],
      "hidden": [],
      "none": [
        "M-0020"
      ],
      "unknown": [
        "M-0014"
      ]
    },
    "io": {
      "shown": [],
      "partial": [],
      "hidden": [],
      "none": [
        "M-0020"
      ],
      "unknown": [
        "M-0014"
      ]
    },
    "training": {
      "shown": [],
      "partial": [],
      "hidden": [],
      "none": [
        "M-0020"
      ],
      "unknown": [
        "M-0014"
      ]
    }
  },
  "implementations": [
    {
      "mechanism": "M-0014",
      "selected": {
        "id": "I-0011",
        "title": "AI 2040 inference-only verification stack",
        "url": "https://trustbutveri.fyi/implementations/ai-2040-inference-only-verification-plan/"
      },
      "implementations": [
        {
          "id": "I-0011",
          "title": "AI 2040 inference-only verification stack",
          "url": "https://trustbutveri.fyi/implementations/ai-2040-inference-only-verification-plan/"
        },
        {
          "id": "I-0010",
          "title": "RAND secure inference data center (SIDC) design",
          "url": "https://trustbutveri.fyi/implementations/rand-secure-inference-data-centers/"
        }
      ]
    },
    {
      "mechanism": "M-0020",
      "selected": null,
      "implementations": []
    }
  ],
  "sources": [
    {
      "id": "S-0067",
      "title": "Verification Plan",
      "authors": "R. Dean",
      "year": 2026,
      "url": "https://ai-2040.com/supplements/verification-plan",
      "path": "/sources/dean-verification-plan/"
    },
    {
      "id": "S-1511",
      "title": "Get Involved in Verification",
      "authors": "AI Futures Project",
      "year": 2026,
      "url": "https://ai-2040.com/supplements/verification-plan/get-involved",
      "path": "/sources/ai-futures-get-involved-verification/"
    },
    {
      "id": "S-1512",
      "title": "Verifying international AI deals: Plan A, the state-of-play, and what you can do to help",
      "authors": "T. Milton et al.",
      "year": 2026,
      "url": "https://amodo.substack.com/p/verifying-international-ai-deals",
      "path": "/sources/milton-verifying-international-ai-deals/"
    },
    {
      "id": "S-1008",
      "title": "AI 2040 Plan A — Verification SITREP",
      "authors": "Amodo Design",
      "year": 2026,
      "url": "https://amododesign.com/ai-verification/plan-a-sitrep/",
      "path": "/sources/amodo-plan-a-verification-sitrep/"
    },
    {
      "id": "S-1312",
      "title": "Fitting a Network TAP to our Inference Verification Prototype",
      "authors": "Amodo Design",
      "year": 2026,
      "url": "https://amododesign.com/notes/2026-09-15-network-tap-inference-verification/",
      "path": "/sources/amodo-network-tap-inference-verification-prototype/"
    },
    {
      "id": "S-0015",
      "title": "Verifying LLM Inference to Detect Model Weight Exfiltration",
      "authors": "R. Rinberg et al.",
      "year": 2025,
      "url": "https://arxiv.org/abs/2511.02620",
      "path": "/sources/rinberg-verifying-llm-inference-weight-exfiltration/"
    },
    {
      "id": "S-1507",
      "title": "Adversarial Entropy Inflation Against Gumbel-Based Inference Verification",
      "authors": "N. Kezins",
      "year": 2026,
      "url": "https://arxiv.org/abs/2608.23375",
      "path": "/sources/kezins-adversarial-entropy-inflation/"
    },
    {
      "id": "S-1314",
      "title": "DiLoCo: Distributed Low-Communication Training of Language Models",
      "authors": "A. Douillard et al.",
      "year": 2024,
      "url": "https://arxiv.org/abs/2311.08105",
      "path": "/sources/douillard-diloco/"
    },
    {
      "id": "S-0060",
      "title": "Does Distributed Training Undermine Compute Governance?",
      "authors": "R. Rahman",
      "year": 2026,
      "url": "https://arxiv.org/abs/2605.29359",
      "path": "/sources/rahman-distributed-training-compute-governance/"
    },
    {
      "id": "S-1301",
      "title": "Traffic Shaping for Workload Classification",
      "authors": "Lucid Computing",
      "year": 2026,
      "url": "https://lucidcomputing.substack.com/p/traffic-shaping-for-workload-classification",
      "path": "/sources/lucid-traffic-shaping-workload-classification/"
    },
    {
      "id": "S-1313",
      "title": "The Tray as a Bandwidth Boundary",
      "authors": "Amodo Design",
      "year": 2026,
      "url": "https://amododesign.com/notes/2026-03-16-dpu-bandwidth-limiter/",
      "path": "/sources/amodo-tray-bandwidth-boundary/"
    },
    {
      "id": "S-1410",
      "title": "Covert AI Projects",
      "authors": "B. Halstead & T. Larsen",
      "year": 2026,
      "url": "https://ai-2040.com/supplements/covert-ai-projects",
      "path": "/sources/halstead-covert-ai-projects/"
    },
    {
      "id": "S-0002",
      "title": "Verifying International Agreements on AI: Six Layers of Verification for Rules on Large-Scale AI Development and Deployment",
      "authors": "M. Baker et al.",
      "year": 2025,
      "url": "https://www.rand.org/pubs/working_papers/WRA4077-1.html",
      "path": "/sources/baker-verifying-international-agreements-ai/"
    },
    {
      "id": "S-1409",
      "title": "Tracking Hyperscale AI Data Center Growth with Satellite Imagery",
      "authors": "C. Krawec",
      "year": 2026,
      "url": "https://fas.org/publication/tracking-hyperscale/",
      "path": "/sources/krawec-tracking-hyperscale-satellite-imagery/"
    },
    {
      "id": "S-1411",
      "title": "Introducing the Frontier Data Centers Hub",
      "authors": "Epoch AI",
      "year": 2025,
      "url": "https://epoch.ai/latest/introducing-the-frontier-data-centers-hub",
      "path": "/sources/epoch-frontier-data-centers-hub/"
    },
    {
      "id": "S-3301",
      "title": "AI Data Centers Documentation – Methodology",
      "authors": "Epoch AI",
      "year": 2026,
      "url": "https://epoch.ai/data/data-centers-documentation/methodology",
      "path": "/sources/epoch-data-centers-methodology/"
    }
  ]
}