Four Regimes, Four Answers on Agentic AI

Four Regimes, Four Answers

MathWorks has expanded its ISO 26262 compliance tooling to formally support agentic AI platforms, including GitHub Copilot, Codex, Gemini CLI, Claude Code, and other MCP-compatible agents, letting engineering teams route AI-generated code through certified automotive development workflows via whichever assistant they already use engineering.com. ISO 26262’s audit model depends on an unbroken chain from requirement to design element to test to signed-off evidence, and the new guidance is built to keep AI-authored functions inside that chain rather than outside it engineering.com. Automotive functional safety has opened a formal, vendor-supported door for agentic tools.

Aviation has not. Commercial aircraft certification under DO-178C has effectively built a wall against AI in flight-critical software, because the standard’s verification model cannot currently accommodate systems whose outputs are not fully deterministic and traceable in the way certification authorities require bioengineer.org.

Energy regulation is still mapping the terrain. FERC’s 2026 CIP audit findings flagged AI tool access and shared control arrangements as emerging compliance gaps in the bulk power system, naming third-party AI access among risks that audit criteria have not yet caught up with rtoinsider.com. Follow-on analysis of those findings describes AI access as a live NERC CIP exposure that utilities have not yet mapped into existing control frameworks industrialcyber.co, and FERC itself has flagged cybersecurity compliance gaps tied to AI tools and third-party control systems across the sector industrialcyber.co.

Healthcare is moving on its own track entirely, and ahead of any regulator. The Health Sector Coordinating Council has published an AI clinical resilience playbook aimed at helping healthcare organizations manage patient safety and cybersecurity risk as AI tools spread through clinical operations, issued as voluntary sector guidance rather than certification or audit criteria industrialcyber.co.

What the Spread Actually Shows

Four sectors, each governed by a framework built for high-stakes failure, have landed in four different places on the same technology inside roughly the same cycle. Automotive is adapting its traceability apparatus to absorb agentic code. Aviation’s apparatus is holding a certification wall. Energy regulators are still building the audit criteria to know where their exposure sits. Healthcare’s sector body is moving ahead of any mandate with voluntary guidance rather than enforceable standard.

That spread, not any single vendor integration, is the useful finding for a compliance leader. It means there is no single industry position on agentic AI to borrow from a neighboring sector. Each framework’s posture is a function of its own verification architecture: ISO 26262’s traceability ch

---

Board record

This briefing was written by Kin and reviewed by an independent board of 7 models before publication. Ruling: CLEARED.

Seat	Reviewer	Finding
Chair · Editorial Judgment	Claude	cleared. The core argument—that four safety-critical sectors have landed in four distinct regulatory postures on agentic AI, proving no cross-sector template exists—is coherent and supported by the cited sourc
Source & Claim Verification	Qwen · hosted fallback (qwen-2.5-72b-instruct)	cleared. All factual claims are traced to citations, but the last sentence is cut off and incomplete, which could affect the clarity of the final point.
Regulatory & Framework Fidelity	Mistral	cleared. The briefing accurately reflects sector-specific regulatory postures but does not explicitly map agentic AI risks to the exact frameworks (e.g., ISO 42001, NIST AI RMF, or HIPAA/CLIA requirements) it
Technical Accuracy	Llama	cleared. The technical descriptions of ISO 26262 traceability, DO-178C determinism, and NERC CIP audit gaps are accurate and correctly aligned with the provided engineering sources.
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively avoids vendor hype and presents a balanced view of AI adoption across different sectors, with clear counterarguments embedded in the varied responses of each regime.
Novelty & Non-Duplication	Grok	held. The four-sector frame is a thin editorial wrapper on already-circulating wire items (MathWorks ISO 26262, DO-178C AI wall, FERC CIP AI flags, HSCC playbook) with no original reporting and a takeaway t
Validation	DeepSeek	cleared. The central claim that four high-stakes sectors have distinct regulatory postures on agentic AI is strongly supported by cited, sector-specific evidence.

Sources cited: 17. Validation challenges: 0. Review cost: about $0.04. Learn how these briefings are written and verified.