Autonomous shipping is often imagined as an empty cargo ship crossing the ocean without human involvement. The reality is more nuanced. Maritime autonomy covers a range of operating modes, from decision-support tools on a conventionally crewed ship to remote-controlled vessels and, in limited cases, ships capable of operating with little or no person on board. The technology is developing rapidly, but safe adoption depends on regulation, communications, cyber security, human oversight and clear responsibility—not only on artificial intelligence.
What is a Maritime Autonomous Surface Ship?
The IMO uses the term Maritime Autonomous Surface Ship, or MASS, for a ship that, to a varying degree, can operate independently of human interaction. “Varying degree” is important. A vessel may have automated steering but still be fully crewed; another may be controlled from a shore-based centre; another may be designed to perform parts of a voyage autonomously inside defined operating limits. These are different risk profiles and should not be discussed as if they were the same technology.
In 2026, IMO adopted a non-mandatory MASS Code that provides goal-based guidance for safe design, construction, operation and certification. It is intended to support safe integration with conventional shipping and create practical experience before a future mandatory code is developed. The code does not mean every commercial ship can now sail without a crew. It means governments and industry have a structured framework for assessing and learning from MASS operations.
The levels of autonomy in practical terms
A useful way to understand autonomy is to ask who makes decisions and where they are made. At the lowest level, the human crew remains in control but uses advanced systems for route optimisation, collision-risk alerts, engine monitoring and adaptive steering. This is already common in modern shipping. At the next level, systems may execute some functions automatically while a crew member supervises and can intervene.
Remote operation moves part of the control task to a shore-based Remote Operations Centre (ROC). The ship may have reduced crew, a safety crew or no crew for a defined period. The shore operator requires reliable communications, clear authority, situational awareness and a safe transfer-of-control procedure. The highest degree describes a vessel that can make and execute decisions without real-time human intervention within its approved operating envelope. Even then, human organisations retain responsibility for design, operating limits, maintenance, emergency response and oversight.
Why autonomous shipping is being developed
Autonomy can improve certain operations. Short, repetitive routes such as ferries, harbour craft, survey vessels, workboats and coastal cargo services may benefit from precise track keeping, predictable schedules and shore support. Uncrewed or remotely operated craft can reduce exposure during dangerous inspections, mine countermeasures, environmental monitoring or offshore work. Automation can also assist conventional crews by reducing routine workload and making abnormal conditions more visible.
Commercial arguments include potential crew-cost savings, improved schedule consistency, fewer accommodation requirements and data-driven maintenance. However, removing crew from a ship does not remove work. It relocates work to software engineers, ROC operators, maintenance teams, cyber-security staff, surveyors, rescue arrangements and shore logistics. The business case must therefore consider the whole system rather than comparing only crew numbers.
The central safety challenge: sensing and understanding the situation
A competent navigator does more than detect another ship. They understand whether it is fishing, constrained, overtaking, manoeuvring unpredictably, carrying out an operation or responding to an emergency. They observe weather, sea state, visual cues, radio traffic, local practice and equipment behaviour. An autonomous system needs sensors—radar, cameras, lidar where appropriate, GNSS, AIS, depth sensors and machinery data—and must combine them into a reliable picture despite rain, fog, glare, false targets, missing AIS data, sensor faults and cyber interference.
Perception alone is not enough. The system must decide whether a situation creates risk, select an action compliant with the COLREGs, communicate a predictable intention and continue monitoring the result. It must also recognise when it has reached the edge of its competence and needs human intervention. This is why the MASS Code stresses risk assessment and operating limits, including limitations related to weather, visibility, water depth, sea state and other conditions.
COLREGs, seamanship and the “human-like” problem
The COLREGs are written in terms such as proper lookout, safe speed, early action and good seamanship. These principles can be applied to automated systems, but engineering them is not a simple translation exercise. A human bridge team can use judgement when another ship gives an unclear signal, when a fishing fleet behaves irregularly or when a local pilot makes a non-standard but safe manoeuvre. An algorithm needs an explicit design response for uncertainty.
Safe autonomous navigation must therefore be conservative enough to prevent collision, but not so cautious that it blocks channels or creates new close-quarters situations. This balance is particularly hard in congested ports, mixed traffic and restricted visibility. Trials must test abnormal and rare events, not only perfect-weather demonstrations.
Remote Operations Centres
A ROC is not simply a control room with screens. It needs trained operators, clear watchkeeping arrangements, suitable staffing, fatigue controls, resilient communications, secure systems, decision support, emergency procedures and an auditable safety-management system. If one operator supervises several vessels, the workload and attention demand must be carefully assessed. A system that performs well in normal traffic may fail when two ships require intervention at once.
Transfer of control is a critical moment. The ship, ROC and any onboard person need a common understanding of who has the conning authority, what mode the vessel is in, what alarms are active, what route or manoeuvre is underway and what happens if communication is lost. These arrangements must be designed, trained and rehearsed just as carefully as a bridge watch handover.
Communications and cyber security
Autonomy increases dependence on data and connectivity. Satellite links, cellular networks, radio systems and onboard networks may all carry operational information. The system must tolerate delay, loss of signal, bad data and deliberate interference. It needs safe fallback modes: slow down, hold position where safe, proceed to a preplanned contingency area, alert a ROC, or transfer control. There can be no assumption that a broadband connection will be perfect throughout an ocean passage.
Cyber security is a navigation and safety issue, not merely an IT issue. A compromised sensor feed, manipulated chart update, disabled alarm or unauthorised remote command can affect the ship directly. Defence includes secure design, access control, software update management, network segmentation, monitoring, incident response and crew/shore awareness. Class societies, flag States and companies increasingly treat cyber resilience as part of safe operation.
Machinery, maintenance and emergency response
A crewed ship handles countless small failures: a leaking pump seal, a blocked filter, a loose cable, a false alarm, a stuck valve or a cargo issue. An autonomous or reduced-crew ship needs another method to detect, diagnose and resolve these events. Condition monitoring, redundancy, remote diagnostics, modular equipment and planned port maintenance become more important. Some failures may require a technician to board at a safe location; others may force the vessel into a degraded but safe operating mode.
Emergency response poses equally difficult questions. Who fights a fire, assists a casualty, responds to flooding or secures dangerous cargo if nobody is on board? How will search and rescue authorities contact and recover the vessel? What equipment, access points and information will responders need? Autonomy should never shift risk from the ship to an unprepared rescue service or nearby mariners.
Jobs will change rather than disappear overnight
The likely employment effect is transformation. Maritime organisations will need remote operators, autonomy supervisors, data analysts, cyber-security specialists, software assurance staff, technicians, surveyors and regulators who understand both technology and seamanship. Deck and engine expertise will remain valuable because it is needed to define safe operating limits, interpret risk and design credible fallback actions. The most resilient maritime careers will combine operational competence with digital literacy.
What happens next?
IMO’s current roadmap envisages an experience-building phase for the non-mandatory MASS Code, followed by development of a mandatory code and related SOLAS work. The timetable and final requirements can change through IMO’s normal intergovernmental process. For now, the sensible expectation is gradual adoption: first in narrow operating domains, then in more complex operations as evidence, regulation, technology and public confidence mature.
Final perspective
Autonomous shipping is not a choice between humans and machines. It is a question of how to allocate tasks while maintaining safety, accountability and resilience. The best systems will not pretend uncertainty has disappeared. They will define their limits, keep humans meaningfully involved, communicate clearly with conventional ships and fail safely when sensors or links are imperfect.