The Certification Wall Is Showing Up in Audit Findings, Not Only Cockpits

Aviation’s resistance to AI in safety-critical functions is not caution, it is math. ARP4754B, ARP4761A, DO-178C, and DO-254 all assume a system whose behavior can be exhaustively verified before failure, and learned, probabilistic behavior does not produce that kind of evidence bioengineer.org. That constraint is well understood in aviation circles. What is newer, and more useful to industrial and utility compliance leaders, is where the same question is now surfacing outside aviation, in audited findings rather than theoretical debate.

The clearest case is the power grid. FERC’s 2026 CIP audit findings flag AI tools and third-party control system access as active compliance risks, alongside gaps in generator registration and shared control arrangements, meaning auditors are already finding AI with a foothold near real-time grid control rather than hypothesizing about it rtoinsider.com. Industrial cybersecurity analysts reviewing the same findings point to the same pattern: AI access and shared control show up as live gaps in NERC CIP compliance, not future scenarios industrialcyber.co. A separate review of the same audit cycle confirms the broader pattern across the US power sector, where AI tools and third-party control system integrations are producing documented compliance gaps under FERC’s cybersecurity oversight industrialcyber.co. This is the aviation question, asked by a different regulator, after the fact instead of before it.

The Factory Floor Version

Manufacturing is getting its own version through two converging tracks. Cobot safety regulation is tightening across Europe, North America, and Asia-Pacific, and certified sensors are shifting from optional upgrade to compliance requirement as collaborative robot deployment grows indexbox.io. Separately, MathWorks has expanded its AI guidance for ISO 26262 compliance to support agentic coding assistants through MCP-compatible workflows engineering.com. That tooling helps teams build and document systems that meet ISO 26262. It does not certify AI as the safety component inside the control loop, and compliance leaders should not let a roadmap slide blur that line.

Edge computing is closing the physical distance between AI workloads and real-time control regardless. OMRON’s acquisition of OnLogic reflects demand for industrial-grade AI computing positioned alongside, not separate from, real-time automation systems ien.eu. Proximity is not admission, but the FERC findings show how quickly the two can be confused in practice once AI tools have network access to operational systems.

The Question That Travels

The grid audits make the aviation framing concrete for industrial buyers. The decision is not whether an AI tool is useful. It is whether that tool has access, shared control, or decision authority inside a certified or NERC CIP-governed control loop, versus sitting outside it accelerating the engineering around that loop. FERC’s auditors already found organizations that had not drawn that line before a regulator did it for them. Industrial and utility compliance teams now have a documented reason to draw it first.

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 aviation’s certification constraints are now appearing as documented audit findings in other regulated sectors—is coherent and well-supported by the FERC audit citations, though
Source & Claim Verification	Qwen · hosted fallback (qwen-2.5-72b-instruct)	cleared. All factual claims are supported by citations, but some sources could be more robust or directly linked to the claims.
Regulatory & Framework Fidelity	Mistral	cleared. The briefing accurately reflects DO-178C and NERC CIP frameworks but does not substantively address ISO 42001, EU AI Act, NIST AI RMF, HIPAA, Cures Act, ONC certification, CLIA, or CMS prior authoriza
Technical Accuracy	Llama	cleared. The text accurately identifies the deterministic verification requirements of aviation standards (ARP4754B, DO-178C, DO-254) and correctly distinguishes between AI-assisted engineering and AI-driven c
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively identifies and counters potential vendor hype by clearly distinguishing between AI’s utility in development versus its role in certified control loops, and by grounding its ar
Novelty & Non-Duplication	Grok	held. The aviation-to-FERC-audit synthesis is a useful cross-domain frame not identical to any single wire piece, but the briefing still leans heavily on very recent, widely duplicated FERC CIP/AI-access co
Validation	DeepSeek	cleared. The central claim that AI is being flagged as a compliance risk in audited systems, not just theoretical debates, is strongly validated by specific, recent FERC audit findings in the power sector.

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