{ "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": "2609.30614", "title": "Subjects, Not Authors: The Authorship Hazard in Agentic Dataspaces", "summary": "The method replaces prompt-stated policy duties and diff-based policy review with compile-time enforcement of ODRL duties at the tool-call boundary and registry-held field classifications routed to human approval. It costs a classification registry (designed but not yet implemented in the prototype), human approval throughput that the authors acknowledge does not scale to the motivating participant volume, and integration with dataspace connectors. It fails where a protected value is not confined to a named field: the compiled condition exposed the value in 7 of 7 such cases.", "rationale": "The core registry mechanism is explicitly not yet implemented, the approval pool is acknowledged not to scale, and the setting is dataspace connector governance rather than mainstream LLM production work, so it merits monitoring rather than adoption.", "technique": "Compiled ODRL constraints as agent governance boundary", "family": "seguranca_e_guardrails", "recommendation": "observar", "published": "2026-09-24", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.30614", "independent_implementations": 0, "total_implementations": 1, "stars": 10, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "The method replaces prompt-stated policy duties and diff-based policy review with compile-time enforcement of ODRL duties at the tool-call boundary and registry-held field classifications routed to human approval. It costs a classification registry (designed but not yet implemented in the prototype), human approval throughput that the authors acknowledge does not scale to the motivating participant volume, and integration with dataspace connectors. It fails where a protected value is not confined to a named field: the compiled condition exposed the value in 7 of 7 such cases.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.30614", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Protected fields reached the model in 0 of 105 cases when an ODRL duty was compiled into an invocation-time tool-call constraint, versus 105 of 105 under prompt-stated duties; unpublished drafts reversed 80 authorization decisions, all caught only by treating field classification as authorship requiring review.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.30614", "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": "" }