Maritime AI Digest — 1 November 2026

This week the news finally talks about the people on board. A study asked 532 seafarers about the new systems they get, and most said the same thing: there is no time to learn them. A robot dog now walks the rounds in a steel mill, and somebody still has to read what it finds. A drone will map the rust on a hull, and a surveyor will still sign. New EU rules say that when software on board hurts someone, the crew is protected and the owner is not. Every story below has a machine in it. Look for the person next to it, and ask one question: did anyone give that person the time to make it work?

The week's most important developments in shipping & oceans — distilled into a 5-minute read.

🔗 Quick Links

• ⏱️ 72% of seafarers say they have no time on board to learn new systems — Splash247

  A study of 532 seafarers in 64 countries. The software makers say good tools need no training. We disagree: serious systems need training, and the missing part is the time to do it

• 🐕 Robot dogs start real inspection rounds, and Equinor is among the first buyers — Splash247

  ANYbotics delivers a four-legged robot that can work where gas may be present. Posco already runs three at a steel mill. No ship uses one yet

• 🚁 Toei Kisen tests AI drones that find rust on a ship's hull — Splash247

  Trials start in November on one of its ships. The hope is dry dock inspection in days, not weeks. Nobody has checked that yet

• ⚖️ If software on board fails, who pays? New EU rules protect the crew, not the owner — Splash247

  From 9 December software counts as a product. Chubb expects AI exclusions in policies, and a Korean insurer will link cyber cover to how secure the ship is

• 🗣️ Google Meet now translates speech live in 63 languages — Android Authority

  Each person hears the call in their own language. Not on dial-in calls, so the ship needs a link that can carry video. No accuracy figures

• 🔐 A free list of security holes in ship equipment, as a gang names 38 maritime firms — Smart Maritime Network

  ClassNK, Bureau Veritas and a cyber firm open the list. Nobody has confirmed the gang's claims, and no ship systems were shown to be hit

• 🧭 AiatSea: every tool this week still needs a person with time — AI at Sea

  A robot needs someone to read its data. A drone needs a surveyor. A translation needs a check. Four questions to ask before anything new comes aboard

🚀 Big Moves This Week

1. 72% of Seafarers Have No Time to Learn New Systems. Training Is Not the Problem. Time Is.

Anyone who has joined a ship in the last five years knows the feeling. A new e-logbook, a new reporting tool and a new performance system, all in the same year. A PDF manual. And a full working day that has not changed at all.

The numbers. A study by the World Maritime University, paid for by the Lloyd's Register Foundation, asked 532 seafarers in 64 countries. 72% said they do not have enough time on board to learn new digital systems. More than 80% get digital skills training rarely or never. Only 13% said training ashore matches the equipment they find at sea. And two out of three want to improve their digital skills. So this is not a crew that refuses to change. It is a crew with no time.

What the software makers said. SplashTech asked people who sell shipping software. Several said the problem is the software, not the crew. "If a tool needs a training course, the tool is the problem," said Adam Dennett of SpecTec. KeelX's Aristos Philis said software should adapt to the crew. NAPA's Mikko Kuosa said the industry buys technology one department at a time, so the crew ends up with five systems and five logins.

Our view, from the ship side. We do not agree that good software needs no training. Simple apps do not. A planned maintenance system, a stability program or an engine monitoring system does, and should. Even a spreadsheet needs training to be used well. Training is a normal cost of any serious system. The real problem is that nobody plans the time for it. The system is bought ashore, installed in port, and the crew learns it between watches. And when it does not work, the crew works around it. One former seafarer, Hrvoje Mihovilović, put it well: "Installation is presented as adoption, cooperation as acceptance."

Be careful with the "no training" message. Every person quoted sells software, and "our tool needs no training" is a sales line. We build software for ships ourselves, so this applies to us too.

What to do. Before you sign for any new system, write down three numbers: how many crew hours the training takes, who on board gets that time, and when. Train on the real ship's equipment, not on slides. Ask the supplier which old task or report this system removes. And after three months, ask the crew, not the supplier, whether it saves time. [Splash247 · AiatSea, 18 October]

2. Robot Dogs Start Real Inspection Rounds. Could One Walk Your Engine Room?

Every engineer knows the round. Walk the engine room, look at the gauges, feel for heat, listen for a noise that was not there yesterday, smell for oil. Now a four-legged robot is being sold to do part of that job.

What happened. The Swiss company ANYbotics has started delivering ANYmal X, a four-legged inspection robot built to work in places where gas may be present, and Equinor is among the first customers. The robot is certified for those areas (ATEX and IECEx, the European and international rules for equipment that could ignite gas). It carries thermal cameras, visual and sound sensors and gas detection. It checks pumps, compressors, valves and pipes for overheating, leaks and strange noises. It climbs stairs and walks narrow walkways without changes to the plant. ANYbotics says it ran hundreds of test missions since 2023 with BP, Shell, Petrobras and others, including offshore trials on the Valhall field in the North Sea. Its software schedules the rounds and sends the findings into the maintenance system.

One that already works. At Posco's steel mill in South Korea, a Boston Dynamics robot dog has checked 30 oxygen ports and hundreds of pipes every hour since October 2024. Before, at least two workers took turns doing it around the clock. Posco runs three of them. Each costs about $200,000. The limits are honest: it works 40 minutes, then rests 20. It cannot close a valve. And a person in a control room still reads everything it finds.

What is missing. No price for ANYmal X. No result from Equinor. And no ship uses one yet.

Why a superintendent should care. An engine room is exactly the kind of place these robots are built for: noisy, hot, full of stairs and pipes, and often with fewer people than before. A robot could do a night round in an unmanned engine room and record the same readings every time. But a ship moves, and these robots were tested on plants that stand still.

What to ask. If anyone offers one for a ship: has it been tested on a rolling deck? How long does the battery last? Who reads the data, and when? And what happens when it finds something it cannot fix? [Splash247 · Korea JoongAng Daily]

3. A Japanese Owner Sends AI Drones to Look for Rust on Its Hull

Toei Kisen, a Japanese shipowner, will test an AI drone system for hull inspection from November, on one of its own ships. The system comes from Exa, a Japanese technology company. It is the first time Exa's system is used on a real ship.

How it works. Drones fly around the outside of the hull and take pictures and 3D measurements. From these, the system builds a digital twin, which means a 3D computer copy of the hull. AI then looks for defects such as rust and marks exactly where each one is on the copy. A third part plans safe flight paths, so the drones do not hit the hull. Exa uses 3D technology first built for the steel maker JFE Steel. The drone company Aeronext helps with the flying, and Forcesteed Robotics with the image analysis. Toei Kisen gives the ship and access to repair yards.

What they hope for. Today a hull inspection in dry dock can take several weeks. The developers hope to bring that down to a few days. Nobody has checked that yet, and the trial is meant to test it in real conditions: different hull shapes, wind, light and old coatings.

What is missing. No class society is named. No ship type or size. No cost.

Why it matters. Dry dock days are expensive, and much of the time goes to people on scaffolding with a torch and a notebook. A drone that marks every rust spot on a 3D copy gives the next dry dock something to compare with. That could be worth more than the time saved.

What to ask. If you are offered drone inspection at your next dry dock: will class accept the findings, or only use them as help? Who owns the 3D copy of your hull afterwards? And can the next yard read it? [Splash247]

4. If Software on Board Fails, Who Pays? New EU Rules Protect the Crew, Not the Owner

Three news items this week answer one question: when software or AI on a ship goes wrong, who pays?

The law. From 9 December 2026, the EU's new Product Liability Directive treats software as a product. If faulty navigation, engine control or autonomy software causes injury, the maker can be liable without anyone proving fault. A court can order the maker to show its evidence. There is no cap on the amount. Security holes can count as a defect too. In a SplashTech article, the legal researcher Farzaneh Javadi points out who benefits. Only people can claim: crew and passengers, for injury and death, and private owners for things like a passenger's car on a ferry. The shipowner cannot claim for the ship, lost hire or salvage. That still depends on the contract with the supplier and on hull insurance.

Two more details. It covers only software put on the market after 9 December. But an update can count as a new product. So the date of the last big update may matter more than the date the system came on board. And suppliers facing unlimited injury claims will try to push the risk back to owners in their contracts.

The insurers. The chief executive of Chubb, one of the world's largest insurers, said on 22 October that AI liability is a growing risk insurers have not yet priced. He expects clearer policy wording first, then exclusions, which means AI losses taken out of normal cover, and then separate AI policies. He did not talk about marine. In Korea, Hanwha General Insurance and the cyber firm CYTUR plan a pilot in December for cyber insurance where the terms depend on how secure the ship is. No prices or details yet.

What to do now. Read any software or AI contract you are renewing with 9 December in mind. Look for clauses where you must protect the supplier against injury claims, and check with your P&I club whether you are covered for that. Ask for access to the system's decision logs and update records. And before your next renewal, ask your P&I club and hull insurers in writing whether a loss linked to AI or software is covered. [Splash247 · Insurance Journal · Digital Ship]

5. Your Crew Call, in Everyone's Own Language

Anyone who has run a safety meeting with a mixed crew knows the moment. The officer explains a new procedure in English. Everyone nods. A week later you find out a third of them did not really follow it. Now picture the same call where each person hears it in their own language, as it is spoken.

What Google launched. On 22 October, Google started rolling out live speech translation in Google Meet, its video call service. It works in 63 languages. Meet detects which language each person speaks, and each listener can hear a translated voice, read translated captions, or both. People can mark the languages they already understand so those are not translated. It works on computers and phones, with Google's business plans from Business Standard up, and its paid personal AI plans. Calls can run up to 8 hours on the top plans.

The limits, and one matters a lot at sea. It does not work for people who join by phone dial-in. It does not work on recordings or live streams. Google gives no accuracy figures, and nothing on how it copes with a weak connection.

Why this is good news. Crewing teams, superintendents and trainers talk to Filipino, Indian, Ukrainian, Indonesian and Chinese seafarers every week. Shipyard meetings in Korea or China often need an interpreter. A tool many offices already pay for may now remove part of that barrier. We want to try it.

Why to be careful. Translation is good at everyday talk and weaker at technical words. "Purifier", "bilge" and "lashing" can come out strange. For a safety instruction, a wrong word is a real risk.

What to try. Test it on one ordinary crew call, with an officer who speaks both languages listening for mistakes. Use it for briefings and questions, not for the final safety instruction. Check that the ship's link can hold a video call. Then ask the crew what they understood, in their own words. [Android Authority]

6. A Free List of Security Holes in Ship Equipment, in the Week a Gang Named 38 Maritime Firms

Last week AI found two old holes in web software that runs most websites. We asked how a ship manager would even know which of their equipment is affected. This week there is a free tool for that question.

The list. The cyber security company Cydome, with ClassNK, Bureau Veritas, MTS-ISAC (a group where shipping companies share cyber warnings) and Rakuten Maritime, has opened the Maritime Vulnerabilities Repository. It is a free public list of known security holes in IT and control equipment used on ships. You can search it by equipment and maker, read short expert notes on some entries, and get an alert when something new appears for your equipment. More than 350,000 security holes have been published in total, with about 200 new ones a day, so finding the few that matter for your fleet is the hard part.

The same week. On 12 October a criminal group listed 38 maritime, shipping, port and logistics companies on its leak site and claimed to have their data. The satellite and IT company Marlink looked into it. It could not confirm that the data is real, and found no proof that ship systems or port operations were hit. Marlink's advice to everyone: check remote access given to suppliers, check file transfers, call your partners, confirm any change to payment or cargo release instructions by phone to a known number, and keep your logs.

What is missing. Cydome sells cyber security, and the article does not say who checks the list long term or how it is paid for.

What to do this week. Make a list of the main equipment on one ship, with maker and software version. Search it in the free list. If the list is too long to make, that is the first finding. [Smart Maritime Network · Digital Ship · AiatSea, 25 October]

7. AiatSea: Every Tool This Week Still Needs a Person With Time

Look at this week's machines again. Each one comes with a person the brochure does not show.

The robot dog at Posco checks 30 ports an hour. A researcher in a control room reads what it finds, and a person still closes the valve. The drone will map every rust spot on a hull. A surveyor still decides what needs repair and signs for it. The translation turns English into Tagalog in real time. An officer who speaks both still has to listen for the wrong word. The free security list names the holes in your equipment. Somebody still has to know what equipment you have. And the new EU law protects the crew when software hurts them, which tells you who is standing closest when it goes wrong.

None of this is bad news. These are useful tools, and we would like to see several of them on ships. But each one moves work. It takes away some walking, climbing or typing, and adds some reading, checking and deciding. The question is who gets the new work, and whether they have time for it.

That is why story 1 matters most. 72% of seafarers already have no time to learn the systems they have. Add a robot, a drone, a translator and a security list, and the time has to come from somewhere. If nobody plans it, the crew finds it at night, or does not find it, and the tool gets the blame.

Our view. Good systems need training, and training needs time. Buying the system is the easy part. Planning the person's time is the real project.

Four questions before anything new comes aboard. Whose work changes on day one, by rank? What work does it remove, and what work does it add? How many hours of training does it need, and when will those hours happen? And who will tell you, honestly, after three months whether it saves time? If the answer to the last question is the supplier, ask the crew as well. [Splash247 · Korea JoongAng Daily · AiatSea, 11 October]

📊 Why It Matters — Strategic Impact Table

🔭 On Our Radar

• 🗳️ Who Is Actually Paying for Shipping's AI Accounts? — carried for a twelfth week: Anthropic's final stock market filing on 26 October puts its future computing commitments at about $518bn and shows two customers bring in about 23% of its revenue, we monitor whether AI suppliers change prices or limits for business customers and whether any ship manager publishes a policy on paid AI tools, and our reader survey is still open (we use Claude, made by Anthropic, to help produce this newsletter).

• ⏱️ Does Anyone Put Training Hours Into the Contract? — 72% of seafarers say they lack time to learn new systems, we monitor whether any owner, manager or union sets a rule for paid training time per new system, and whether any supplier publishes the crew hours its system needs.

• 🐕 Does a Robot Dog Board a Ship? — inspection robots now work in plants and offshore, we monitor whether any owner tests one in an engine room at sea, and what happens to battery, footing and data on a rolling deck.

• 🚁 What Does Toei Kisen's Drone Trial Show? — the trial starts in November, we monitor how long the inspection takes compared with the old way, whether class accepts the findings, and who keeps the 3D model of the hull.

• ⚖️ Do Supplier Contracts Change Before 9 December? — EU product rules now cover software, we monitor whether software suppliers add new indemnity clauses for owners, whether P&I clubs warn members about them, and whether the first AI exclusion appears in a marine policy.

• 🔐 Does the Free Security List Grow? — the list is open to new contributors, we monitor which equipment makers and class societies join, whether any maritime supplier confirms its status on last week's web software holes, and whether any of the 38 named firms confirms a breach.

• 🖥️ Can Ships Run Big AI Without the Link? — Mistral has now released its large AI model for anyone to download and run on their own computers, we monitor whether any maritime supplier uses such models on board, offline, and on what hardware.

• 🇰🇷 Korea Now Aims for Ships With No Crew on Board by 2032 — a new government plan also targets 30% less damage from marine accidents by 2030, we monitor whether the permanent test area off Geoje is approved this year and whether any results are published.

📅 Critical Maritime AI Research Areas for Managers

1. ⏱️ Crew Time Needed for New Systems: most seafarers lack time to learn new software. Research should measure the real crew hours each common system needs to learn and use, and compare that with the time it saves after six months.
2. 🐕 Inspection Robots on a Moving Ship: robot dogs now patrol plants that stand still. Research should test battery life, footing, sensor accuracy and data handling in a real engine room at sea, in different sea states.
3. 🚁 Drone Hull Inspection and Class Acceptance: drones can map rust on a 3D model of the hull. Research should compare drone findings with a surveyor's findings on the same hull, and set out what class would need to accept them.
4. ⚖️ Software Liability in Ship Contracts: new EU rules make software makers liable for injury without proof of fault. Research should review current supplier contracts for indemnity and update clauses, and publish model wording that protects owners.
5. 🗣️ Live Translation for Crew Communication: live translation is now built into common video tools. Research should test accuracy on ship technical words and safety instructions in the main crew languages, over real ship connections.
6. 🔐 Equipment Lists for Cyber Checks: free lists of security holes only help if owners know what is on board. Research should find the simplest way for a mid-size manager to keep an up-to-date list of equipment and software versions per ship.
7. 🧭 Who Gets the New Work: every new tool removes some tasks and adds others. Research should track, by rank, which tasks change when a new system goes live, and whether the work added is planned or simply absorbed.

📈 Top Investment Opportunities

1. ⏱️ Crew Training That Fits the Ship — 72% of seafarers lack time to learn new systems and only 13% say shore training matches the equipment at sea — the opportunity is in short training built on each ship's own equipment, done on board in small pieces, and in tools that measure the hours each system really needs — Splash247 · WMU study
2. 🐕 Inspection Robots for Hazardous and Unmanned Spaces — explosion-proof robot dogs are now delivered to energy companies — the opportunity is in versions tested for ships and offshore units, and in the software that turns their readings into maintenance work — ANYbotics · Equinor
3. 🚁 Drone Inspection and 3D Hull Records — an owner is testing AI drones that map rust on a 3D copy of the hull — the opportunity is in inspection services that class can accept, and in hull records that pass from one dry dock to the next — Toei Kisen · Exa
4. ⚖️ Insurance and Records for Software Risk — EU law now treats software as a product and insurers expect AI exclusions — the opportunity is in software liability cover for maritime suppliers, cyber cover linked to measured security, and tools that keep version and decision records — EU rules · Splash247
5. 🔐 Equipment Lists and Security Checks for Fleets — a free list of security holes in ship equipment is now open — the opportunity is in services that keep each ship's equipment list up to date and match it against new security holes automatically — Blue Initiative

📅 Top Monthly Picks

1. 🧭 One Person Fewer on the Bridge, and the Officer Is Overloaded Seven Times as Often — an ABS computer model showed overload rising from 8% to 61% of port approaches when one person left the bridge; this week 72% of seafarers said they lack time to learn new systems, which is the same problem seen from the training side — Digital Ship
2. 🥽 Smart Glasses Can Show the Procedure and Hide the Pipe — people wearing one headset hit their heads about twice as often in a mock steering gear room; this week Lloyd's Register's technology chief said tools on board should give time back to the crew, not take their attention — Splash247
3. 🏭 Factory AI Agents That Read Alarms and Maintenance History, on Site — agents built for plants read the same data an engine room produces; this week robot dogs started the rounds that collect that data, in plants first — Automation News
4. 📞 CMA CGM Puts AI on the Phone, and Almost Half the Calls Never Reach a Person — AI agents close 46% of cargo release calls; this week live translation arrived in video calls, another sign that AI is changing how shipping people talk to each other — Digital Ship
5. 🧮 Two Hundred Software Flaws or Twenty, the Risk Came Out Almost the Same — a study found ship systems with very different numbers of known flaws scored almost the same risk; this week a free list of security holes opened, and the same company now works with an insurer on cyber cover — Digital Ship

Have feedback or a story tip? Email us or connect on LinkedIn.