IN Brief:
- CK Line and developer Control F signed an agreement to test vessel risk prediction on the 2,693-TEU Sky Pride.
- The proposed system will combine video, weather and ship-motion information to identify potentially hazardous maritime conditions.
- The government-supported project has not yet published detection accuracy, incident reductions or a wider deployment date.
CK Line has agreed to test an artificial intelligence system for detecting hazardous conditions and assessing navigation risks on one of its working containerships. The South Korean operator has signed an AI adoption and cooperation agreement with domestic developer Control F, selecting the 2,693-TEU Sky Pride as a trial vessel. Data collected during normal voyages will be used to develop and evaluate the risk-prediction models; no completed test results have been published.
The planned system combines voyage video, weather information and ship-motion data to identify potentially hazardous conditions, including sea fog, and assess risks associated with vessel movement. Its outputs are intended to feed a monitoring interface for shipboard or shore-based personnel. The companies have not specified a commissioning date, trial duration, sensor configuration or accuracy threshold. Signing an adoption agreement therefore establishes a development project, not a deployed safety service with measured performance.
Sky Pride is a Panama-flagged full container ship built in September 2025, according to CK Line’s own vessel register. The operator lists a design capacity of 2,693 TEU, deadweight of 35,276 tonnes and 340 reefer plugs. Those particulars identify a working commercial ship rather than an isolated test platform. They do not establish that every onboard camera or motion sensor will be integrated with the new system; the detailed data collection arrangement has yet to be published.
Each data stream describes a different part of the voyage. Cameras may show a change in visibility or an apparent obstruction; weather observations provide environmental context; and motion records show how the vessel is responding to manoeuvres and sea conditions. To combine them meaningfully, readings need consistent timestamps, sufficient coverage and a dependable connection to the vessel’s current operating state. Control F has not disclosed its model architecture, video processing method or how conflicting measurements will be handled.
Detecting fog is also different from calculating the risk of a close-quarters encounter. A visibility warning indicates that a condition has developed; an assessment of a particular navigation hazard would need additional information about traffic, relative movement and the ship’s situation. Neither company claims that the prototype generates collision-avoidance manoeuvres, controls the helm or replaces the bridge watch. Its declared role is to support monitoring and risk assessment, with any wider navigation function requiring its own development and verification.
Performance testing will have to account for substantial variation in the images and movement data collected at sea. Glare, darkness, precipitation and contamination on a camera lens can affect visibility, while vessel loading, wave conditions and turns alter motion readings. Weather forecasts may also differ from conditions at the ship’s position. The original agreement does not specify the training dataset or how many different voyage conditions will be sampled, so the intended model’s reliability cannot yet be quantified.
A warning system must balance missed hazards against alerts that consume bridge officers’ attention without identifying a material risk. Evaluating the prototype would require recorded observations, a defined method of assessing correct and incorrect detections and a way for personnel to examine why an alert appeared. The companies have not published any such evaluation results. Claims of fewer incidents, reduced collision risk or faster decisions would be premature until the proposed shipboard trial yields evidence.
The work is being undertaken through a South Korean government-supported programme that encourages established companies to adopt AI developed by domestic start-ups. Oversight involves the Ministry of Science and ICT and the National IT Industry Promotion Agency. Control F’s experience includes models built from weather and environmental data. The vessel project extends that capability into moving-camera imagery, ship behaviour and operational safety decisions, which will need to be tested under actual voyage conditions.
CK Line carries container and bulk cargo on routes connecting South Korea with other Asian markets. Sky Pride appears in its published fleet register, which lists 16 ships. Testing aboard an operational vessel allows developers to observe changing conditions during scheduled voyages, but brings constraints of access to equipment, crew workload and communications availability. The agreement does not establish whether processing will take place aboard ship, ashore or through a combination of the two.
The immediate milestones are installation arrangements, collection of usable voyage data and a documented assessment of the warnings produced. The parties have not announced an incident-reduction result, a certified collision-avoidance capability or a timetable for fitting the system to additional ships. Until those results exist, the project remains a trial of risk-assessment technology aboard Sky Pride, with officers and established navigation procedures still responsible for the vessel’s operation.
