{
  "schema_version": 1,
  "builder_plan": {
    "scenario": "An agent reads external documents or tool responses that may contain malicious instructions.",
    "change": "Replay attacks and legitimate lookalikes in an isolated copy of your agent workflow.",
    "baseline": "Your existing checks with the same tasks, permissions, and tool responses.",
    "steps": [
      "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.",
      "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.",
      "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.",
      "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test."
    ],
    "measure": "Successful attacks missed, legitimate actions blocked, detection delay, and review cost.",
    "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation."
  },
  "builder_review": null,
  "arxiv_id": "2610.23362",
  "title": "Monitor-Validity Failures: LLM Safety Evidence Paths at Evolving Framework Boundaries",
  "summary": "The method replaces informal trust that safety monitors' detector outputs reach enforcement points with fixed canary contracts: confirmation units that place a known payload on a changed evidence surface and record both monitor visibility and the downstream decision. It costs only small deterministic test fixtures and negative controls wired into CI against schemas, streams, tools, and adapters, with negligible runtime overhead since the tests run on framework upgrades rather than in the serving path. It can fail to catch failures outside the canaried surfaces, give false assurance if canaries are not updated when frameworks add new evidence paths, and its findings transfer only partially to closed or third-party-hosted frameworks whose internals the reader cannot instrument.",
  "rationale": "The reader runs LLM applications behind runtime monitors whose framework dependencies evolve, and these low-cost regression contracts directly address a confirmed class of silent safety-evidence failures without requiring scale, clusters, or model retraining.",
  "technique": "Canary contracts for monitor evidence-path regression testing",
  "family": "seguranca_e_guardrails",
  "recommendation": "adotar",
  "published": "2026-10-16",
  "as_of": "2026-10-19",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2610.23362",
  "independent_implementations": 0,
  "total_implementations": 1,
  "stars": 1573,
  "citations": null,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "The method replaces informal trust that safety monitors' detector outputs reach enforcement points with fixed canary contracts: confirmation units that place a known payload on a changed evidence surface and record both monitor visibility and the downstream decision. It costs only small deterministic test fixtures and negative controls wired into CI against schemas, streams, tools, and adapters, with negligible runtime overhead since the tests run on framework upgrades rather than in the serving path. It can fail to catch failures outside the canaried surfaces, give false assurance if canaries are not updated when frameworks add new evidence paths, and its findings transfer only partially to closed or third-party-hosted frameworks whose internals the reader cannot instrument.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.23362",
      "source_page": null,
      "source_excerpt": ""
    }
  ],
  "exposure_map": [
    {
      "dimension": "quality",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "compute",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "latency",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "operations",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "compatibility",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "security",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "data_and_training",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "reproducibility",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    }
  ],
  "risks": [],
  "minimum_test": [],
  "open_questions": [],
  "independent_tests": [],
  "lab_reproductions": [],
  "report_available": false,
  "validation_tier": "unknown",
  "evidence_tier": "unknown",
  "infrastructure_basis": "unknown",
  "equations": [],
  "equations_status": "not_fetched",
  "core_kind": "",
  "equations_fetched_at": "",
  "equations_source_url": "",
  "equations_source_sha256": "",
  "exposures_reviewed_at": "",
  "exposures_source_sha256": ""
}