ShipPaulJobs Weekly AI Brief, November Week 2, 2026: Ports Wait for AI Returns, Windward Adds Cameras, and the AI Trade Wobbles

AI in shipping, ports and shipyards, and the AI industry news that will reach them, as of 7 November 2026

Captain Paul

Maritime 4.0 · AI, Data & Cyber Security · LinkedIn

Ask a terminal manager today whether the port uses AI, and the answer is usually yes. Ask what that AI has added to the bottom line, and the room goes quiet. That gap between using AI and earning from it ran through this week's news. A port survey put a number on it. Maritime AI companies kept buying and launching anyway. A Korean shipyard found a way to get class societies to trust an AI inspector. And on Wall Street a $2.5 trillion selloff asked the same question about the whole AI industry, on the very day Maersk said AI is helping to fill its ships. This is the first ShipPaulJobs Weekly AI Brief. Each week it takes a handful of stories and explains what is going on underneath them, and what they mean for people who run ships, ports and yards.

At a Glance

Date	Where	What	Status
3 & 6 Nov 2026	Global ports	BCG and TIC4.0 survey: only 4% of container port respondents report measurable earnings gains from AI	Survey, reported by press
5 Nov 2026	Global	Windward acquires Azimut.ai to add camera-based vessel identification to its maritime AI platform	Company announcement
5 Nov 2026	Greece / global	AskChief opens its AI tool for technical and compliance teams to the wider market	Company announcement
29 Oct 2026	Korea	HD Hyundai Heavy Industries' AI coating inspector; MOU with LR, ABS, DNV and KR on AI inspection procedures	Company announcement, reported by press
5 Nov 2026	USA / global	A $2.5 trillion tech selloff questions the AI buildout that is lifting container demand	Market data, reported by press
Now → 2 Dec 2026	EU	EU AI Act: what already applies, and the next date for generative-AI marking	Regulation in force

Part 1 · Maritime AI

Ⅰ. Ports: AI Is Spreading, Returns Are Not

Picture a container terminal that has spent two years on AI. One model reads bills of lading and customs papers so clerks no longer retype them. Another watches the cranes and warns before a gearbox fails. The pilots worked, people like them, and now the board asks a simple question: how much money has this made us? A survey published this week suggests most terminals cannot answer it.

Boston Consulting Group and the Terminal Industry Committee 4.0 (TIC4.0) asked 103 experts from terminal operators, port authorities, shipping lines and technology vendors on five continents in August 2026. The results were reported by FreightWaves on 3 November and by WorldCargo News on 6 November. AI is clearly past the experiment stage: 44% said their organisation has reached the scaling stage or later. But only 4% reported measurable earnings gains.

Why such a gap? The survey itself points to the first reason. Only 11% of respondents systematically track what AI does to their efficiency or margins. A terminal that does not measure cannot report a gain, even if one exists. The other reasons are familiar to anyone who has tried to put software into an operation: when asked about barriers, 41% named an unclear return on investment, 32% named data quality and 30% named integration with existing systems. An AI model can only be as good as the data it is fed, and in a port that data sits in a terminal operating system, crane controllers and paper that were never designed to talk to each other.

Where AI is earning its place is less glamorous than the headlines. The two leading use cases are extracting data from shipping documents, which 62% are exploring or implementing, and predicting failures of cranes and straddle carriers, at 48%. The benefits people cite follow the same pattern: operational efficiency (74%) and faster decisions (44%) come well ahead of cost reductions (41%) and customer experience (18%). There is one more detail worth noticing. Unclear return was named by 44% of staff below the C-suite but by only 21% of senior leaders.

Captain Paul's reading

The number I would hang on the wall is 11%, not 4%. If only one in nine organisations measures what AI does for it, the 4% says as much about measurement as about results. And when the people closest to the work are twice as doubtful as the people who approved the budget, that is usually a sign the measuring needs to start on the quay, not in the boardroom.

Ⅱ. Windward Buys Azimut.ai: From What Ships Say to What Cameras See

Most of what the world knows about where a ship is comes from the ship itself. The Automatic Identification System (AIS) is a transponder that broadcasts a vessel's identity, position and course to anyone listening. It is, in effect, a self-report. A ship that does not want to be seen can switch it off or broadcast a false position, and that weakness sits at the heart of the sanctions-evasion problem often called the dark fleet.

That is the background to an acquisition Windward announced on 5 November. The maritime AI company has bought Azimut.ai, a maritime visual intelligence company, for an undisclosed price. Azimut.ai uses optical sensing to identify and "fingerprint" vessels in real time at chokepoints and ports, including vessels whose AIS is off. Windward says it will place cameras at key chokepoints and combine what they see with the ownership, sanctions, cargo and behavioural data it already holds.

A simple way to think about it: AIS is the boarding pass a passenger shows; the camera is the one at the gate. Windward's customers will not receive raw images or video, and the company says the system analyses vessel activity, not individuals. CEO Ami Daniel called the deal "our first step in building a proprietary visual intelligence layer".

Captain Paul's reading

For ordinary operators the practical effect arrives through compliance. When a charterer or bank screens a counterparty, it will increasingly have a second, harder-to-fake signal next to AIS. A vessel with a gap in its AIS track may now also have, or lack, a camera sighting that explains the gap.

Ⅲ. AskChief Opens to the Market: AI That Reads the Company's Own Manuals

Think of a superintendent answering a vetting observation late in the evening, or a chief engineer chasing an alarm on a sister ship's purifier. They rarely lack knowledge. What they lack is time to find the right page: the maker's manual, the procedure in the safety management system (SMS), the record of the last time the same thing happened. AskChief's co-founder Vassilis Mavrogiannis put it this way: "The bottleneck is getting the right evidence in front of the right person".

AskChief, an AI platform for ship technical and HSQE teams, opened to the wider market on 5 November after nearly a year of live use with shipowners and managers. It has two applications, AskChief Technical and AskChief Compliance. Both answer questions from a company's own technical manuals, SMS, vessel records and past cases, and both show the sources behind each answer so staff can check them. Seanergy Maritime, Castor Ships, Iblea, Delia and IMECS (part of the ITOCHU Group) worked with the company on the tools.

In plain words

A general-purpose chatbot answers from what it learned during training, which can be out of date or simply wrong about your ship. A tool that answers from your own documents works more like a fast librarian: it finds the relevant pages first, then writes the answer from them. If it shows which page it used, a person can check the answer in seconds instead of trusting it.

Captain Paul's reading

Showing the source is the right design, because on a ship an unverifiable answer is worse than no answer. The question an operator should ask before rolling out any tool like this is where its manuals and SMS will be stored and processed, and who can see them. Those documents describe how the company's ships are built and run, which is exactly what an attacker would like to read.

Ⅳ. Shipyards: HD Hyundai Brings the Class Societies In

Coating inspection is one of the oldest crafts in a shipyard. An inspector walks a block, looks closely at the paint film and decides whether a defect needs repair. It depends heavily on the individual's eye and experience, and the yard's result is later looked at again by the owner's and the class society's surveyors.

HD Hyundai Heavy Industries announced on 29 October an AI system for this job, AI CIP (AI Coating InsPector). It uses vision AI trained on large sets of coating-quality data from the group's own yards to find defects, replacing judgement that depended on the individual inspector. Development began in 2024 with HD KSOE's AX Center, and the company says the system has passed technical review by major classification societies. No accuracy figures were published.

The more interesting part came the same day. HD Hyundai Heavy Industries, HD KSOE and HD Hyundai Samho signed an MOU with Lloyd's Register, ABS, DNV and Korean Register. The yards will apply the system and improve how inspection data is collected; the class societies will support technical review and work on standard inspection procedures for AI-based inspection.

Captain Paul's reading

Why bring the class societies in? Because in shipbuilding an inspection only counts when the people who sign the certificates accept it. A yard can use any tool it likes for internal quality control, but for AI to replace part of a survey, the surveyor has to be able to trust and audit it. The open questions are about evidence: how the model's decision is recorded, and how a disputed result is settled between yard, owner and class.

Part 2 · The AI Industry, and What It Means for Shipping

Ⅴ. A $2.5 Trillion Selloff Meets AI Cargo

In shipping we usually talk about AI as software, on board or in the office. This week was a reminder that it is also cargo. A data centre needs servers, chips, transformers and power equipment, and a great deal of that moves by sea. So when investors start doubting the AI buildout, it is not only a technology story.

On 5 November the Nasdaq fell 5.5%, about $2.5 trillion in market value and its biggest one-day percentage drop since April 2025 according to Reuters, as reported by Gizmodo. The Philadelphia semiconductor index had its worst day since early 2020. Reports linked the fall to weak AI-sector earnings and to the US midterm results, with investors expecting tougher scrutiny of the AI industry. UBS estimates, per Reuters, that about 30% of the US data-centre capacity expansion expected in 2028 may be at risk of construction pauses.

On the same day, Maersk raised its 2026 profit guidance for the third time, to an underlying EBITDA of $14–14.5 billion, on the back of higher freight rates rather than more cargo. Its shares still fell 11%. CEO Vincent Clerc said demand is "increasingly" driven by investment in electrification, energy and AI infrastructure, which generates significant industrial cargo flows.

Captain Paul's reading

Put the two stories side by side and the link is plain. Part of today's demand for ships rests on a building programme that markets are now questioning. If even part of the data-centre pipeline pauses, as UBS warns, project and breakbulk cargo tied to it softens. Owners and charterers exposed to that trade should watch AI spending numbers as closely as freight indices.

Ⅵ. Rules: What the EU AI Act Already Requires

Recent headlines said the EU had delayed its AI Act. That is true for only part of it. The Digital Omnibus on AI, Regulation (EU) 2026/1744, moved the dates for high-risk systems but did not pause the rest, according to a 5 November summary and law-firm analysis. It helps to think of the Act as a fleet delivery schedule: some ships are already in service, one is due next month, and the largest have had their delivery dates pushed back.

Already in service: the bans on prohibited practices and the duty to give staff adequate AI literacy (since 2 February 2025); the obligations on general-purpose AI models (since 2 August 2025); and most of the Article 50 transparency duties, such as telling people they are talking to a chatbot and labelling deepfakes (since 2 August 2026).

Due next: on 2 December 2026, machine-readable marking of AI-generated content (Article 50(2)) applies to generative systems already on the market before 2 August 2026, and new bans take effect on AI that generates non-consensual intimate images or child sexual abuse material.

Delayed: stand-alone high-risk systems (Annex III) move to 2 December 2027, and high-risk AI built into regulated products (Annex I) to 2 August 2028. Commentators describe the Article 50(2) timing slightly differently, so check the Official Journal text before relying on a date.

In plain words: where marine equipment fits

The delayed Annex I date does not work the same way for every product. Annex I of the Act lists existing EU product laws in two groups. For products under the laws in Section A, such as machinery, the AI Act's high-risk requirements apply directly. The Marine Equipment Directive (2014/90/EU) is in Section B, alongside the laws for civil aviation, vehicles and rail. For Section B products, Article 2(2) says that only a few articles of the AI Act apply. Instead, Article 105 amended the Marine Equipment Directive itself: when the Commission sets technical specifications and testing standards for AI systems that are safety components, it must take into account the high-risk requirements in Chapter III, Section 2 of the AI Act. Put simply, a maker of type-approved marine equipment will meet these AI rules through future MED specifications and testing standards, not through a separate AI Act certificate. (Source: Regulation (EU) 2024/1689 on EUR-Lex, Article 2(2), Article 105 and Annex I.)

What are those requirements? Chapter III, Section 2 is the core of the high-risk rules: a risk management system, data and data governance, technical documentation, record-keeping (logs), information for the people using the system, human oversight, and accuracy, robustness and cybersecurity. Whichever route the rules take to a product, the AI Act directly or the Marine Equipment Directive, the questions will be of the same kind: show how this system was tested, on what data, how often it is wrong, and how a person can step in when it is.

Captain Paul's reading

For a shipping company in the EU, the duties that bite now are the unglamorous ones: staff who use AI tools need adequate AI literacy, and a customer-facing chatbot must say it is a chatbot. For equipment makers, neither the delay nor the MED route is a reason to wait. The evidence those rules will ask for, what data a model was trained and tested on, how accurate it was, how it fails and how a person overrides it, can only be collected while the system is being built. A file put together in 2028 cannot describe tests that nobody wrote down in 2026. And as the HD Hyundai MOU shows, yards and class societies are already asking for this kind of evidence before any law requires it.

The week in one paragraph

AI in shipping is moving from "can it work?" to "can we show it pays?", and most organisations cannot yet answer the second question because they are not measuring. The tools that look most likely to pay are the practical ones: reading documents, predicting failures, finding the right page in a manual, spotting a ship that has gone dark. For AI to count in the formal world of ships, a surveyor, a regulator or a bank has to be able to check it, which is why the class MOU and the EU rules matter as much as the technology. And AI now moves freight markets too.

Ⅶ. What to Watch Next Week

1. Whether the AI selloff spills into data-centre project delays, and what that means for project and breakbulk bookings.
2. More detail from the BCG and TIC4.0 report on how terminals measure AI value.
3. Any class society guidance on accepting AI-based inspection results, following the HD Hyundai MOU.
4. EU AI Act: preparation for the 2 December 2026 marking date for existing generative AI tools.
5. If you use AI on company documents: check where the data is processed, and that answers cite their sources.

Sources

EUR-Lex: Regulation (EU) 2024/1689 (Artificial Intelligence Act), Article 2(2), Article 105 and Annex I

• FreightWaves: New survey finds AI rush isn't boosting ports' earnings (3 Nov 2026)
• WorldCargo News: Only 4% of container port operators report AI earnings gains (6 Nov 2026)
• Cyprus Shipping News: Windward expands into maritime visual intelligence with acquisition of Azimut.ai (5 Nov 2026)
• Smart Maritime Network: AskChief opens maritime AI platform to wider market (5 Nov 2026)
• SBS Biz: HD현대중공업, 선박 도장 결함 AI로 잡아낸다 (29 Oct 2026)
• Gizmodo: Tech rout wipes out $2.5 trillion (5 Nov 2026)
• CNBC Africa / Reuters: Maersk raises outlook again as AI buildout, port congestion keep freight rates high (5 Nov 2026)
• Encorp.ai: EU AI Act after the Digital Omnibus: what applies now (5 Nov 2026)
• Gibson Dunn: EU AI Act Omnibus agreement, postponed high-risk deadlines and other key changes

Captain Paul

Maritime 4.0 · AI, Data & Cyber Security

Writes on AI, data and cyber security for the shipping and shipbuilding industries. Publishes a weekly brief on AI in shipping, ports and shipyards, and on the AI industry news that will reach them.

🤖 All AI Brief issues ↗

⚓ Join the ShipPaulJobs Community

Join →