Six years after its deployment, SatShipAI, an operational platform combining AI models with Sentinel-1 SAR imagery and AIS data, has provided robust maritime surveillance around Denmark. A meta-analysis of archived outputs, logs, and manual reviews shows stable vessel detection and classification performance over time, including successful cross-sensor application to X-band SAR data without retraining. Key operational challenges included orbit file delays, nearshore detection limits, and emerging infrastructure such as wind farms. The platform proved particularly valuable for detecting offshore “dark” vessels beyond AIS coverage, informing maritime security, traffic management, and emergency response. These findings demonstrate the feasibility, resilience, and adaptability of long-term AI–geospatial systems, offering practical guidance for future autonomous monitoring infrastructure.

The rapid proliferation of tracking sensors—ranging from vessel and vehicle tracking systems to smartwatches, cameras, and Earth observation sensors—has led to an unprecedented influx of high-frequency, high-volume data. Yet, despite this abundance, many trajectories remain incomplete, contain errors, or are entirely missing. A vast reservoir of tracking data remains unexplored or underutilized, holding valuable insights that could enhance monitoring and decision-making. The MMDEC dataset aims to serve as a resource for both teaching and the research community, and is specifically designed to support advanced maritime trajectory analysis and inference. Its core contribution lies in the explicit integration of SAR-based, AI-driven ship detections and attribute estimates with AIS vessel reports and multiple contextual data layers, enabling direct comparison between transmitted vessel positions and independently observed radar targets. MMDEC aggregates and harmonizes multi-source data including AIS streams, satellite imagery, meteorological and oceanographic fields, port locations, and sea state and marine protected area boundaries over western Celtic Sea, the English Channel, and part of the North Sea. The dataset covers a continuous three-month period from 1 July to 30 September 2023. Its multimodal content and explicit spatial–temporal alignment provide a multi-sensor benchmark for evaluating algorithms designed to enhance, validate, or infer maritime trajectories from heterogeneous observational sources.