🔔 Subscribe to ShipUniverse Weekly →
Maritime intelligence company Windward announced the acquisition of Azimut.ai on November 3, 2026, adding shore-based and shipborne camera technology capable of detecting, visually recognizing and tracking vessels even when they are not transmitting AIS signals. Azimut reports that an earlier pilot at Israel's Port of Ashdod detected and documented 464 targets, with vessel detection at distances of up to 7,000 meters using standard daylight cameras. Windward plans to combine optical observations with its existing vessel ownership, sanctions, cargo and behavioral intelligence. Financial terms were not disclosed. The technology offers an additional way to verify vessel movements at monitored ports and chokepoints, although camera visibility and reliable vessel identification remain distinct technical challenges.
Turning Off AIS No Longer Guarantees a Ship Goes Unseen
Windward's Azimut.ai acquisition adds fixed-camera and shipborne observations to vessel tracking. The system aims to identify physical ships and connect sightings with movement history, ownership and sanctions records.
Targets detected and documented, supplier-reported
Vessel detection with standard daylight cameras
False alarms after filtering in the pilot
No transaction valuation or global camera count released
Records vessel appearance and movement inside its field of view, independently of an AIS transmission.
Hull and movement may still be observed.
AIS SIGNAL ABSENTRecognize a vessel in the camera's usable field of view.
Estimate position, direction and speed from video observations.
Check visual features against available vessel records.
Combine AIS, satellite, RF and movement history where available.
Evaluate ownership, sanctions and relevant behavior with context.
Direct optical observations at monitored locations, a further source of vessel movement evidence, and technology that can integrate with existing cameras.
Global continuous tracking, universal identification of AIS-silent vessels, independently audited accuracy across diverse ports, or a disclosed network rollout schedule.
Five Data Layers. One Vessel Identity Problem.
A ship's presence, movement, claimed identity and legal ownership are separate questions. Adding visual recognition can help reconcile them, but each source still has blind spots.
A camera records a ship crossing a monitored entrance without a corresponding AIS signal. Analysts can investigate the gap, but absence alone does not prove evasion.
A transmitted track appears inconsistent with physical observation. Investigators compare time, position, vessel features and independent sensor data.
A ship limits transmissions in a high-threat area for safety. Camera evidence may help verify movement without treating the captain's decision as misconduct.
Two ships share hull proportions or visual markings. The system suggests a match, but analysts need independent identity evidence before taking adverse action.
Targets detected during the trial. Not necessarily 464 unique ships, confirmed IMO identities or AIS-dark vessels.
Reported using standard daylight cameras. Does not establish identical recognition accuracy at the maximum distance.
Azimut claims correct identification of innocent vessels during evaluation. This is not an audited overall vessel-identification success rate.
Existing port cameras may provide infrastructure for additional analytics. The acquired platform could help customers reconcile port entries, vessel movements, sanctions screening and historical tracking gaps.
Camera placement needs suitable access, ownership transparency and permissions. Windward says customers receive vessel analytics rather than raw images or video. Data-use practices and identity disputes still require clear operating procedures.
How Many Dark Vessels Could a Camera Network Identify?
Model the portion of AIS-silent traffic that cameras might detect, how many observations receive a tentative vessel identity, and the analyst workload required to review them.
The full bar represents all modeled AIS-silent passages. Green observations receive a tentative identity; blue are seen but not matched; red remain outside successful visual detection.
Baseline number of movements needing other evidence.
Detected with or without an exact candidate identity.
Potential matching leads, not legally verified identities.
Spurious alerts from the assumed station network.
Matched observations, unresolved sightings and false alerts.
Staff time assuming the selected review time per alert.
Station operating charges plus analyst review time.
Not the price of a verified sanctions finding.
Includes unseen passages and unresolved visual sightings.
Adjust the assumptions to see how the system performs and what limits identification coverage.
We welcome your feedback, suggestions, corrections, and ideas for enhancements.
Please click here to get in touch