The third ‘De-mystifying digitalisation and AI in maritime’ roundtable series was held at the offices of Squire Patton Boggs in Singapore, during Singapore Maritime Week, with a focus on risk and liability. The result was a fascinating exchange of ideas, with huge value offered by each of the 17 participants, including our incredible featured speakers: Vicky Koo and Daisy Radford.
Risk and liability in maritime AI
When a class society executive, a shipowner-turned-CFO, port innovation technologists, maritime lawyers, insurers and communications strategists sit down at the same table, the conversation about artificial intelligence stops being theoretical. That was the premise of a full-day, closed-door roundtable held on the sidelines of Singapore Maritime Week, which was part of Intent Communications’ Demystifying Digitalisation and AI in Maritime roundtable series. Convened to interrogate where AI is genuinely changing maritime operations, where the industry is getting ahead of itself, and where regulation, liability and trust are struggling to keep pace, the closed-door format allowed for a genuinely candid exchange.
Over two sessions, participants worked through four connected issues: the reliability of AI outputs, cybersecurity and data sovereignty, liability and the regulatory response, and governance of human oversight. Each surfaced the same tension — between the operational upside AI offers and an industry not yet equipped, culturally or structurally, to absorb it safely. What follows is a summary of the issues raised, and where the room saw the path forward.
Here are some of the key takeaways from the day:
Issue 1: Can the industry trust what AI tells it?
The morning session opened with a warning rather than a welcome. One participant, deliberately invited for their scepticism, set the tone: large language models produce ‘near-miss’ errors that are more dangerous than obvious failures, because they can quietly steer an organisation’s strategy in the wrong direction without anyone noticing. Others agreed that AI performs well on mundane, high-frequency tasks such as data extraction, but becomes genuinely risky when the people using it lack the baseline knowledge to challenge its output. One attendee described how their organisation now requires explicit disclosure whenever AI tools are used in preparing a report, treating transparency as a first line of defence.
Recommendation: For high-stakes use cases, participants argued that companies should move away from open-ended, generative AI and towards retrieval against curated, controlled datasets, thus trading flexibility for auditability. Several also pointed to the need for structured onboarding of any new AI tool, rather than allowing it to enter workflows informally.
Issue 2: Connectivity is outpacing security
The second issue was cybersecurity, sharpened by the industry’s own success in getting vessels online. Ship-to-shore connectivity via services such as Starlink was flagged as having opened a genuinely new threat surface at sea, and one participant noted that some shipowners are already removing Starlink terminals over concerns about where their data physically resides. Several attendees observed that this was an echo of European enterprises pulling back from US cloud providers over data sovereignty. The group also discussed the rise of ‘zero-click’ attacks, citing a widely reported case in which simply opening WhatsApp was enough to compromise a device, with no action required from the victim at all.
Recommendation: The consensus was that awareness training alone cannot address threats of this sophistication. Technical safeguards that are built into systems rather than relying on human vigilance were repeatedly described as the more credible defence. More broadly, the group argued the industry
needs a trust architecture spanning the whole ecosystem, not just individual companies, if it is to guard against deepfakes and social engineering as convincingly as it now guards against conventional intrusion.
Issue 2: Liability has not caught up with capability
The afternoon session moved from technical risk to legal exposure. Delegates heard that the IMO’s Maritime Safety Committee will consider, at its meeting in May, a new agenda item on performance standards for AI embedded in marine equipment, which is a proposal co-sponsored by much of the EU, targeting AI operating quietly inside subsystems such as radio communications and bridge equipment, often without seafarers realising they are interacting with it. The room was divided on whether a seven to
eight year IMO regulatory cycle can keep pace with AI moving fast enough that, as one speaker put it, a single chip-based model can already run tens of thousands of times faster than mainstream chatbots, and whether prescriptive standards risk the “scope creep” seen elsewhere in maritime digital standards.
The liability debate was sharper still. Where does responsibility sit when AI is a decision-making tool rather than a simple assistant? Participants raised the uncomfortable possibility that seafarers could face personal criminalisation for accidents involving automated systems they did not design and cannot fully audit, while the providers of those tools typically disclaim all responsibility in their terms of use.
Recommendation: Several participants pointed to Bureau Veritas’s NI 692 guidelines and to emerging ‘traffic light’ risk registries that categorise AI use cases from green (low risk, minimal oversight) to red (high risk, requiring certified human sign-off) as pragmatic bridges while formal regulation catches up. The unresolved question, repeated more than once, was how jurisdictions can harmonise divergent regulatory approaches for the multinational companies operating across both.
Issue 4: What does “human in the loop” actually mean?
The final thread tied the day together: governance and human oversight. Legal teams, participants agreed, need clear structures so that junior lawyers remain accountable even as AI accelerates their work. This echoed a wider concern that today’s efficiency gains are quietly eroding the fundamental learning that produces tomorrow’s experts. The room also questioned whether current AI training and education efforts are substantive or merely ‘tick box’ exercises, with several arguing that automation bias such as when crew or staff quietly defer to a system rather than performing manual checks, is already a live risk.
Recommendation: Rather than settling on a single definition, the group agreed the industry must be explicit about what ‘human in the loop’ is actually for: verification, or simply the assignment of liability. Participants also discussed how thinly that human oversight can be spread (one captain per ship, or one captain remotely monitoring several) before it stops meaning anything at all.
Where this leaves the industry
No roundtable of this kind produces a final answer, and this one did not try to. What it produced instead was a shared vocabulary for a fast-moving problem: AI in shipping is not one conversation but several, being held onshore and at sea, assistive and decision-making. The industry’s regulatory, legal and cultural frameworks are being asked to keep pace with all of them at once.
Three questions were left deliberately open: how liability will be managed when a tool’s provider disclaims all responsibility; how jurisdictions can harmonise divergent regulatory approaches; and how the industry balances the demand for connectivity against its growing exposure to cyber risk.
Those questions are likely to define the next stage of this conversation at Intent Communications’ inaugural Demystifying AI in Maritime Summit, taking place in London on 6 October. The summit will bring the same cross-sector rigour, convening class societies, shipowners, insurers, lawyers, technologists and regulators in the same room, to a wider audience, and will build directly on the issues raised across this roundtable series.
With thanks
This roundtable was made possible by the generous support of our platinum sponsors, Bureau Veritas and Maritime Passport; our cobalt sponsor, Britannia P&I Club; and our bronze sponsors, Stream Marine Training and Intelligent Seas. We are also grateful to Squire Patton Boggs for hosting the roundtable and generously providing their office space in Singapore. Their commitment to advancing this conversation is greatly appreciated, as is the candour and expertise of everyone who joined us in the room.
Like this article? Click here to sign up for our free monthly ‘De-Mystifying Digitalisation and AI in Maritime’ newsletter and get the latest cross-sector expert insights, news, interviews, videos and case studies straight to your inbox. You can also buy a ticket to the Demystifying AI in Maritime 2026 London Summit here.