“We have decades of experience or common standards for air defense, maritime awareness, data formats. The question is whether we apply that same rigor to AI before the technology outpaces the frameworks, or after,” Maj. Gen. Paul Lynch said at the US Geospatial Intelligence Foundation’s annual GEOINT Symposium in Denver.
Maj. Gen. Paul Lynch’s warning at the GEOINT Symposium
Maj. Gen. Paul Lynch, a British Royal Marine serving as NATO deputy assistant secretary general for intelligence, told the GEOINT Symposium audience that the rapid adoption of AI in imagery analysis, change detection and multisource fusion is reshaping what geospatial intelligence (GEOINT) can deliver — and is creating urgent interoperability challenges for NATO. He said the alliance’s response will be decided “in the next three years.”
Lynch framed the issue bluntly: while AI is “reducing the time from collection to actionable product and enabling analysts to focus on tasks that require human judgment,” it also multiplies governance questions because each of NATO’s 32 members develops its own policies, rules and regulations for AI use and for sharing derived products.
AI-enabled GEOINT and the risk of contradictory national reports
Lynch outlined a practical interoperability problem: two allied nations could each produce intelligence with their own “national AI model” trained on their own imagery data sets, labeling conventions and analytical priorities, then deliver conflicting reports to the same NATO commander. “Which one does the commander use, on what basis, with what confidence?” he asked, calling this the “AI interoperability challenge for allied GEOINT.”
His point: when the data feeding decisions has passed through distinct, undocumented models with different assumptions, commanders lack a common baseline for confidence and attribution — a shortcoming that could distort operational decision-making.
Commercial GEOINT, classification, and legacy frameworks
Lynch warned that NATO’s existing processes were built for a different world and are straining under the influx of commercial satellite constellations and industry-processed GEOINT. He noted the alliance already struggles to incorporate commercial imagery into military and intelligence systems in ways that promote member-state interoperability.
GEOINT, he explained, is chiefly about providing location and change-detection information about human activities and natural phenomena such as wildfires, using satellite imagery, maps and other data. But those commercial data and products now must be fused with national imagery, open-source and partner-provided intelligence and then delivered across “32 national classification systems and a set of legal and contractual frameworks that were written for most of those capabilities existed,” Lynch said. He added sardonically that this reality “means it’s not” straightforward.
He also said commercial data currently “enters NATO through intelligence systems, mostly through exceptions and workarounds, not designed pathways,” highlighting ad-hoc practices that leave integration fragile and inconsistent.
Needed standards: metadata, model documentation, attribution and confidence
For Lynch, the operational problem has a governance solution. He called for agreed standards on how AI models are trained and documented, how AI-enabled products are attributed, and what confidence thresholds are operationally usable in particular contexts. He urged “common meta-data schemes, common AI model documentation, [and] common interfaces that don’t require bespoke integration every time a new partner or new source joins the enterprise.”
He framed the path to an “AI enabled, allied intelligence advantage” as running “primarily through governance, not necessarily through additional capability,” emphasizing processes, documentation and shared rules over purely technical upgrades.
What this means for NATO commanders, national analysts, and commercial providers
- NATO commanders: Expect to face intelligence products derived from heterogeneous AI pipelines with differing provenance and confidence statements unless common thresholds and attribution practices are adopted; decisions will depend on agreements about which products are operationally trusted.
- National analysts and security policy teams: Will need to document model training data, labeling conventions and analytical assumptions so other allies can interpret and fuse outputs without misreading intent or masking biases.
- Commercial GEOINT providers and contract officers: Must anticipate tighter data-use policies, security classification guidance and releasability rules following NATO’s June adoption of its first commercial space strategy — work Lynch described as the “unglamorous” but necessary development of data use and contract frameworks.
Lynch closed by restating the urgency: NATO’s frameworks for incorporating commercial and AI-processed GEOINT must evolve quickly or risk being overtaken by the technology they are meant to control. The alliance has acknowledged the commercial space era by signing a commercial space strategy, but the unglamorous work of data policies, classification guides, contract frameworks and releasability rules — complicated further by AI — now dominates the agenda. The practical question he left on the table is stark and time-bound: will NATO put common standards in place before AI changes the operational calculus forever, or after?
