TAK 5.6 is the most capable version ever shipped
TAK 5.6 released in March 2026 with performance improvements, enhanced collaboration, and expanded plugin support. The free ATAK drone plugin lets operators fly drones through ATAK with real-time position on the map. COTAK hit 5,000 registered users as of January 2026. The UK and Finland ran a 3,000-soldier joint exercise using TAK. Malaysian Air Force SOF adopted it via GoTAK.
TAK is winning. It is the de facto common operating picture for dismounted operations across NATO and partner forces. Infra-TAK enables single-command TAKServer deployment. TAK-Portal prototypes TAK as a Service. The adoption curve is steep and accelerating.
TAK sees everything but understands nothing
TAK is a visualization and communication layer. It shows icons on a map, shares Cursor-on-Target messages, displays video feeds, routes chat. TAK does not process sensor data, classify threats, prioritize information, or generate intelligence products. A camera feed appears on the ATAK display. Whether that feed contains a person, a vehicle, or nothing of interest is a question TAK cannot answer.
By design. TAK is a protocol and display framework, not an AI platform. The TAK Product Center has AI integration on the roadmap for late 2026 or 2027. As of June 2026, there is no native AI inference in any TAK client. No object detection. No threat classification. No automated alerting. The operator watches every feed and makes every classification call manually.
The integration layer already exists
EdgeLance publishes Cursor-on-Target XML into TAK networks. When local AI detects a person, vehicle, or object of interest, it generates a CoT entity with position, classification, confidence, and timestamp. That entity appears on every ATAK display as a standard marker. No custom plugin. No TAK modification. Standard CoT protocol over standard multicast or TAKServer relay.
The integration is bidirectional. EdgeLance receives CoT markers from other TAK sources, incorporating them into the local threat picture. An ATAK user who drops a point of interest generates a CoT message that EdgeLance ingests, logs, and can query against. The TAK service layer handles translation, routing, and deconfliction. Shipping code, not a roadmap item.
What AI-enhanced TAK actually requires
Adding AI to TAK is not a plugin problem. It needs local inference so it works when SATCOM is down. Model management so operators run approved, current models. Evidence capture so detections have provenance. Mesh networking so intelligence reaches operators off the TAKServer network. Compute policy so inference runs within device hardware constraints.
TAK's architecture does not support these requirements natively, and it should not have to. TAK is a display and communication protocol. The AI layer is a separate system that feeds into TAK through CoT, the same way a radio feeds through its gateway or a camera feeds through a video server. EdgeLance provides the complete node-level AI stack that TAK needs as an upstream producer. The five-thousand-dollar ISR stack running on a MacBook, connected to a camera, publishing detections as CoT, is a force multiplier every TAK user could deploy.