The Navigation Layer Nobody Is Certifying

The Navigation Layer Nobody Is Certifying

Every recent conversation about autonomous flight certification centers on the same object: the AI system making decisions in the cockpit or the flight computer. Regulators want to know how confident they can be in its outputs. Insurers want to know how to price its failure modes. Global Aerospace’s recent briefing on regulatory and insurance approaches to autonomous flight systems frames the problem almost entirely this way, as a question of what the AI decided and why.

That framing skips a layer. Autonomous and AI-assisted flight systems don’t reason in a vacuum. They depend on positioning data, and that data is not guaranteed. Ground Control’s selection into the DNK GNSS resilience tracking program exists precisely because GNSS signals for drones and robotics are subject to degradation, spoofing, and denial, conditions that no certification framework currently treats as a first-class input to an autonomy safety case. If the navigation layer feeding an AI decision system is compromised, the quality of the AI’s reasoning becomes irrelevant. The output is only as trustworthy as the input it was built on.

This matters because the assurance models now emerging for aviation AI, including the UK Military Aviation Authority’s outcome-focused approach of treating AI development like any other system engineering discipline and requiring assurance to a defined confidence level, are built around the assumption that inputs are stable enough to evaluate the decision logic on its own terms. GNSS resilience programs suggest that assumption is shakier than the certification conversation acknowledges. A confidence level assigned to an autonomy algorithm under nominal GPS conditions tells a regulator or underwriter very little about behavior under degraded or denied positioning, which is precisely the scenario most likely to produce a claim.

The UN panel warning that current AI guardrails are unraveling points to the same structural gap from a different angle. Panel member Qinghua Lu noted that aviation built its safety culture on incident reporting, independent scrutiny, and layered safeguards, but questioned whether those practices scale to more capable and more interdependent systems. GNSS denial is exactly the kind of layered, environmental failure mode that incident reporting regimes were designed to catch. If autonomous flight certification treats navigation integrity as a solved input rather than a monitored variable, the layered safeguard model breaks at the seam between infrastructure and algorithm.

For compliance and safety leaders evaluating autonomous or AI-assisted flight programs under FAA or EASA pathways, the practical implication is straightforward. A certification or insurance submission that documents AI decision confidence without documenting positioning resilience under adversarial or degraded GNSS conditions has an incomplete safety case, regardless of how rigorous the AI assurance work looks on paper.

The industry is starting to test GNSS resilience operationally, ahead of any regulatory requirement to do so. That sequencing should be a signal, not a comfort.

---

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 certifying AI decision systems without certifying their positioning inputs creates an incomplete safety case—is logically sound and well-constructed, though the leap from a sing
Source & Claim Verification	Qwen · local	cleared. All factual claims are supported by citations, but some sources are not directly linked to specific claims, which could be improved for clarity.
Regulatory & Framework Fidelity	Mistral	held. seat error: Client error ‘404 Not Found’ for url ‘https://openrouter.ai/api/v1/chat/completions’
For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/404
Technical Accuracy	Llama	cleared. The article correctly highlights the importance of considering GNSS resilience in autonomous flight certification, a crucial aspect often overlooked in current discussions.
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively identifies and counters a potential oversight in AI certification discussions by highlighting the critical, often unaddressed, issue of GNSS resilience, without resorting to v
Novelty & Non-Duplication	Grok	cleared. The GNSS-as-uncertified-input framing is a genuine synthesis not present in any single cited wire story, though it remains an incremental repackaging of two concurrent items rather than a catalogue-br
Validation	DeepSeek	cleared. The central claim that certification frameworks inadequately address GNSS resilience as a primary input to AI safety is strongly supported by cited industry and regulatory sources acknowledging the vu

Sources cited: 11. Validation challenges: 0. Review cost: about $0.04. Learn how these briefings are written and verified.