https://ond.cc/news Research, field writing, and product notes on governing AI agent actions before they run. en Mon, 19 Oct 2026 15:36:24 GMT https://ond.cc/blog/matoi-reuse-evidence-recheck-authority https://ond.cc/blog/matoi-reuse-evidence-recheck-authority Fri, 16 Oct 2026 04:00:00 GMT Research What remains valid when an agent's environment changes? Matoi studies how to reuse checked consequence evidence without treating it as permission to execute. The research preprint is now on arXiv. https://ond.cc/research/matoi https://ond.cc/research/matoi Thu, 15 Oct 2026 04:00:00 GMT Research Reusing checked consequence evidence as an agent's environment changes, while keeping current authority, grants, quota and uncertain outcomes as separate obligations. https://ond.cc/blog/cava-pcaa-independent-research-citations https://ond.cc/blog/cava-pcaa-independent-research-citations Sun, 04 Oct 2026 04:00:00 GMT Company Two independent agent-security preprints cite our work on action identity and approval binding. A small milestone worth celebrating, with the sources in full view. https://ond.cc/blog/egress-is-an-action-boundary https://ond.cc/blog/egress-is-an-action-boundary Tue, 29 Sep 2026 04:00:00 GMT Research Exact-action grants contained the observed scope violations that destination filtering missed. The same study exposes where revocation, single use and completion guarantees end. https://ond.cc/blog/action-pass-suica-agent-governance https://ond.cc/blog/action-pass-suica-agent-governance Tue, 22 Sep 2026 04:00:00 GMT Research Inspired by Suica and FeliCa, we tested what happens when exact-action approval is prepared before dispatch. The whitebook, arXiv paper and reproducible evidence are now available. https://ond.cc/research/zerogate https://ond.cc/research/zerogate Mon, 21 Sep 2026 04:00:00 GMT Research Separating exact-action approval from durable one-time admission, with explicit revalidation, preparation costs, and replayable evidence. https://ond.cc/blog/agent-registries-are-not-action-proof https://ond.cc/blog/agent-registries-are-not-action-proof Fri, 04 Sep 2026 04:00:00 GMT Perspectives Microsoft Agent 365 is a useful signal for where the market is going. But enterprise agent registries and lifecycle control planes still need an action-boundary proof layer underneath real side effects. https://ond.cc/blog/babyblue-v11-registry-scoped-run https://ond.cc/blog/babyblue-v11-registry-scoped-run Wed, 19 Aug 2026 04:00:00 GMT Engineering OSuite and Baby Blue / invinoveritas ran a v11 registry-scoped reference run: OSuite created a CAVA action artifact, received a signed independent-mediator verdict, self-submitted it to Baby Blue's ledger, executed a real GitHub issue creation, and checked that the outcome still matched the approved action. https://ond.cc/blog/harness-runtime-governance https://ond.cc/blog/harness-runtime-governance Wed, 12 Aug 2026 04:00:00 GMT Engineering AI agent harnesses help models plan, call tools, and run longer workflows. Enterprise runtime governance answers a different question: what action was approved, what actually executed, and can the proof survive audit? https://ond.cc/blog/auto-mode-needs-runtime-governance https://ond.cc/blog/auto-mode-needs-runtime-governance Sat, 08 Aug 2026 04:00:00 GMT Engineering As coding agents move toward longer-running autonomous work, the enterprise control point shifts from whether a model may call a tool to whether the runtime action stays inside an approved boundary. https://ond.cc/blog/cava-benchmark-computable https://ond.cc/blog/cava-benchmark-computable Sat, 08 Aug 2026 04:00:00 GMT Research A public replication survey initially marked CAVA's benchmark claim as not released. We published the missing reproducibility artifact, and the record was updated to COMPUTABLE after an independent rerun. https://ond.cc/blog/external-verifier-checkpoints https://ond.cc/blog/external-verifier-checkpoints Thu, 30 Jul 2026 04:00:00 GMT Engineering OSuite and Baby Blue / invinoveritas ran a concrete external-verifier composition: a CAVA action artifact, a partner-signed verdict, and a proof packet that can be checked without trusting either party's internal log. https://ond.cc/blog/runtime-boundary-benchmark https://ond.cc/blog/runtime-boundary-benchmark Wed, 29 Jul 2026 04:00:00 GMT Research OSuite Runtime Boundary Benchmark tests whether an agent action is still inside the boundary that was approved, across 20 runtime surfaces, 38 risk families, 6 locales, and 10 obfuscation styles. https://ond.cc/blog/runtime-coverage-is-not-a-parser-number https://ond.cc/blog/runtime-coverage-is-not-a-parser-number Tue, 28 Jul 2026 04:00:00 GMT Engineering CAVA gives OSuite a stable action language. Runtime Coverage explains how each real agent runtime is mapped into that language. https://ond.cc/blog/final-authority-arxiv https://ond.cc/blog/final-authority-arxiv Thu, 16 Jul 2026 04:00:00 GMT Research The paper argues that frontier providers can shape model-side safety, but enterprises still need final authority over the actions taken inside their own environments. https://ond.cc/blog/database-log-is-not-agent-governance https://ond.cc/blog/database-log-is-not-agent-governance Thu, 16 Jul 2026 04:00:00 GMT Perspectives Recording what happened is useful. It is not the same as deciding what is allowed before an AI agent changes something that matters. https://ond.cc/blog/spreadsheet-cannot-classify-agent-consequence https://ond.cc/blog/spreadsheet-cannot-classify-agent-consequence Thu, 16 Jul 2026 04:00:00 GMT Perspectives An Excel sheet can track agent activity. It cannot reliably turn messy runtime behavior into a governable action object. https://ond.cc/blog/approval-should-expire https://ond.cc/blog/approval-should-expire Thu, 16 Jul 2026 04:00:00 GMT Perspectives If an AI agent can reuse approval without a boundary, the approval becomes a standing permission. BAF exists to stop that drift. https://ond.cc/blog/dashboard-is-not-runtime-exposure-graph https://ond.cc/blog/dashboard-is-not-runtime-exposure-graph Thu, 16 Jul 2026 04:00:00 GMT Perspectives A dashboard shows metrics. AREG shows how agent risk can travel across runtimes, tools, systems, approvals, and proof gaps. https://ond.cc/research/cava https://ond.cc/research/cava Thu, 16 Jul 2026 04:00:00 GMT Research A runtime-semantics layer for turning shell commands, MCP calls, SDK calls, workflow steps, and gateway requests into canonical action objects that OSuite can score, route, approve, attest, and prove. https://ond.cc/research/final-authority https://ond.cc/research/final-authority Thu, 16 Jul 2026 04:00:00 GMT Research A governance argument for deployer sovereignty: frontier providers should not be the permanent final authority over every downstream enterprise decision. https://ond.cc/research/runtime-exposure-management https://ond.cc/research/runtime-exposure-management Fri, 03 Jul 2026 04:00:00 GMT Research The customer-facing operating layer that turns governed agent actions into exposure inventory, remediation backlog, dependency risk, snapshots, and security evidence. https://ond.cc/research/agent-runtime-exposure-graph https://ond.cc/research/agent-runtime-exposure-graph Sat, 27 Jun 2026 04:00:00 GMT Research A graph model for runtime exposure evidence and uncertainty, with research-only reachability bounds over supplied occurrences rather than complete future-program closure. https://ond.cc/research/bounded-action-firewall https://ond.cc/research/bounded-action-firewall Fri, 26 Jun 2026 04:00:00 GMT Research Bounded authorization for a specific action and execution context, with strict context/effect matching and a proposed connection to state-dependent exposure evidence. https://ond.cc/research/pcaa https://ond.cc/research/pcaa Wed, 24 Jun 2026 04:00:00 GMT Research A deployer-side governance primitive for binding human or policy approval to the action an agent will actually execute, rather than to a mutable prompt, wrapper call, or after-the-fact log. https://ond.cc/research/decision-score-v21 https://ond.cc/research/decision-score-v21 Tue, 16 Jun 2026 04:00:00 GMT Research The scoring model used to explain why a governed action was auto-approved, escalated, blocked, or routed to a human reviewer. https://ond.cc/research/policy-to-runtime https://ond.cc/research/policy-to-runtime Fri, 12 Jun 2026 04:00:00 GMT Research How OSuite maps governance frameworks, internal policy profiles, and runtime agent behavior into one approval and proof loop.