The International Maritime Organization’s (IMO) comprehensive review of the STCW Convention and Code presents a rare opportunity to put artificial intelligence literacy at the centre of maritime training standards.
AI is advancing across shipping much faster than existing training frameworks can adapt. The current STCW review could therefore become one of the industry’s most important opportunities in years to establish a common foundation for the safe and responsible use of these technologies at sea.
Artificial intelligence is already being applied across maritime operations. It can analyse machinery data, identify maintenance trends, support weather routing, optimise fuel consumption and help crews navigate complex regulations. Used appropriately, these tools can reduce workloads and deliver important information to seafarers more quickly.
But there is an important distinction between fluent output and sound judgement. An AI system can produce a detailed, confident and convincing answer while still being completely wrong.
That is why AI literacy should be formally recognised as an STCW competence. Seafarers do not need to become data scientists, but they do need to understand how to identify fabricated information, biased outputs, missing context and data-security risks. Most importantly, they must understand when professional judgement and direct observation should take precedence over an automated recommendation.
AI is already at sea
I am an early adopter and a strong supporter of these technologies, but their weaknesses are anything but theoretical when applied onboard a vessel.
A poor AI recommendation can affect the safety of the crew, the ship, its cargo and the marine environment. The question is no longer whether seafarers will use AI. They already are, either directly or through increasingly intelligent equipment and shore-based services.
The real question is whether they are prepared to use these systems responsibly.
Consider a voyage recommendation built from weather conditions, vessel traffic and fuel-consumption data. Before acting on it, an officer should ask whether the information is current, whether all relevant operational constraints have been taken into account and whether the recommendation is consistent with what is being observed from the bridge.
The same approach applies to predictive-maintenance alerts, cargo planning and regulatory searches.
Knowing how to write a prompt or navigate an AI interface can be useful, but neither should be mistaken for competence. True AI literacy means understanding a system’s limitations, checking its output against the operational context and remaining accountable for the decision that follows.
That is particularly important because AI frequently communicates with an air of certainty.
Seafarers are already trained to question colleagues, verify calculations and compare information from multiple sources. They should have the same professional confidence when questioning a machine, particularly when its response appears polished, immediate and authoritative.
A rare opportunity for STCW
The IMO began its comprehensive STCW review in 2024 and identified more than 400 potential gaps during the first phase. Amendments are now being developed, with completion targeted for 2029–2030.
The scale and timetable of this process highlight why AI literacy should not be treated as an optional addition or left entirely to individual shipping companies and training providers.
A revised framework should establish an international baseline.
Everyone working at sea should have a basic understanding of AI, while officers and personnel responsible for AI-enabled systems should receive deeper training in validation, limitations, escalation procedures and override decisions.
The standard should define what competent behaviour looks like without becoming tied to a particular product or platform that could become obsolete within a short period.
The enduring competence is not mastery of one system. It is the ability to ask the right questions, verify evidence, protect sensitive information and recognise when experience or direct observation should outweigh an automated recommendation.
Those principles are likely to remain relevant even as AI technology continues to evolve throughout a seafarer’s career.
Formal standards, however, will need to be reinforced through continuous learning. Digital training can help keep knowledge current between formal courses and provide short, relevant scenarios to seafarers wherever they are working.
AI can also be used to personalise that learning. But education must ultimately promote curiosity and professional judgement rather than passive acceptance of machine-generated answers.
If the current STCW review fails to address AI literacy, the shipping industry could have to wait years for another opportunity with a comparable international reach.
By then, AI-assisted decision-making is likely to be even more deeply embedded in everyday maritime operations. The safeguards should therefore be established now, while there is still time to shape how these technologies are used at sea.
What seafarers need to learn
Seafarers must understand that generative AI can invent facts, references and explanations. Any safety-critical claim produced by an AI system should be checked against approved manuals, established procedures, charts, regulations and trusted operational data.
They also need to recognise the risks created by bias and missing information. An AI system may understate an unusual local condition, rely on an incomplete dataset or optimise an outcome that is not actually the right operational objective.
A competent user should therefore ask what evidence supports the answer, what information may be missing and what the consequences would be if the recommendation were wrong.
Data protection is equally critical. Crews need to understand what information must never be entered into a public or unauthorised AI platform. This includes personal information, commercial data, ship vulnerabilities and protected operational records.
Convenience should never create a new pathway for data leakage or cyber risk.
Most importantly, AI advice must never become operational authority.
A system can help draft a report, investigate a problem or compare possible options. It cannot assume professional responsibility. That responsibility remains with the officer, who must be capable of explaining the evidence, challenging the recommendation and deciding when it should be rejected.
Assessment should focus on these behaviours rather than simply testing definitions.
A practical scenario, for example, could present a false citation, incomplete data or an apparently efficient plan that conflicts with a safety limitation. The seafarer should be expected to identify the weakness, seek confirmation, communicate uncertainty and select the safer course of action.
That is the practical competence the STCW framework now needs to recognise.
