Your crew are already using AI. On a rig, a platform or a vessel, that comes with two problems most onshore businesses never face: the connection can vanish at any moment, and every question typed into a public AI tool sends your data off the site.


Picture 03:00 on a supply vessel in the North Sea. A pump alarm has gone off, and the second engineer wants the fault-finding procedure. The manual is a PDF that runs to hundreds of pages. The satellite link is up, so he does what millions of people do at work every day. He pastes the fault history and a section of the maintenance log into a free AI chatbot and asks what to check first.

He gets an answer. He also sends your maintenance data to a server you don’t control, in a country you didn’t choose, under terms nobody on board has read.

Now imagine the link drops halfway through. He has no answer, and the data has already gone.

That’s the offshore AI problem in one scene. Here’s why it’s getting harder to ignore.

The problem: AI got offshore before the rules did

Your crew already use it, approved or not

In October 2025, Microsoft research found that 71% of UK employees had used unapproved consumer AI tools at work, and 51% did so every week. Offshore crews, on long shifts and far from the people who know the answers, have every reason to do the same.

The UK’s National Cyber Security Centre (NCSC) calls this “shadow AI”. It warns that staff who move sensitive information into consumer AI services reduce the organisation's visibility and control over it, because that information may be stored, kept or used to improve the service.

Every prompt is data leaving the site

Permits to work, lift plans, incident reports, maintenance histories and crew details are exactly the kind of thing people paste into a chatbot to tidy up or summarise.

The NCSC’s advice is blunt: don't include sensitive information in queries to public AI models, and don’t ask anything that would cause problems if it were made public. The provider can see and store what you type, and stored queries can be hacked, leaked or made public by accident.

“Do loose prompts sink ships?”

the NCSC’s own subtitle for its guidance on AI chatbots

Offshore assets are a target

This isn’t only about careless data handling. Energy and maritime infrastructure is a deliberate target.

In February 2024, US agencies, joined by the UK’s NCSC and other partners, warned that a Chinese state-sponsored group known as Volt Typhoon had compromised critical infrastructure organisations, mainly in communications, energy, transport and water. They assessed it was pre-positioning on IT networks so it could move on to operational systems and disrupt them in a crisis.

At sea, the trend is the same. A Cytur report found that maritime cyberattacks doubled in 2025. Voyage logs, cargo manifests, ship design schematics and crew personal data were found for sale online. One common attack encrypts a ship’s planned maintenance system and holds the voyage logs to ransom.

It’s also about to get faster. In its annual review this month, the NCSC said it is “highly likely” that AI-powered malware will have been used in a ransomware attack on UK critical national infrastructure by 2028. Energy and transport are both on that list.

Each connection between your site and an outside service is one more way in, and one more way out.

The link can’t be relied on

Cloud AI assumes a good, constant connection. Offshore, you can’t promise one.

On 24 July 2025, Starlink had a global outage lasting about two and a half hours, caused by a failure in its core network software. Users in every region lost their connection at the same time. ThousandEyes’ advice to network teams afterwards was to plan for exactly this: name your critical service dependencies and set out how you’ll work in a degraded or offline state.

Even when the link is up, every document uploaded to a cloud AI tool competes for bandwidth with the traffic that keeps the asset running.

Experience is getting thinner

The people asking the questions are changing too. BIMCO and the International Chamber of Shipping estimate a shortage of 39,100 certified officers in 2026. That means more crew who need the right procedure quickly, on the night shift, with nobody experienced to ask.

The regulators are already asking the question

None of the main offshore and maritime cyber rules mention AI by name. All of them ask what is connected, what can get in and what can get out.

A public AI tool on a crew laptop is an outside connection that probably isn’t in anyone’s risk assessment.

Banning it won’t work

The obvious response is to block AI tools. The NCSC’s view is that shadow AI is unlikely to disappear completely, and that organisations should give people approved, secure alternatives that meet their needs. Block the tools without offering anything better, and people find a workaround.

The NCSC has also pointed to the way through. For sensitive organisational data, it says self-hosted models may be appropriate after a proper security assessment.

Your crew want AI. Your data shouldn’t leave the site. You can have both.

The solution: Echo7, on board

Echo7 Offshore puts a complete AI assistant on your vessel, rig or platform. It runs on board, on your own hardware and inside your own network. Your documents, knowledge base and search stay on site, and so does the AI model that reads them and answers questions, rather than sitting in someone else’s data centre.

Because the whole AI chain is on board, it works with no internet connection at all, and nothing leaves the site.

What crew can do with it

Crew ask questions in plain language on the vessel network and get answers from your own material, with the source quoted, so they can check it:

  • operating procedures and OEM manuals
  • HSE documents, method statements and permits to work
  • shift handovers and maintenance history
  • reports, checklists and documents they upload to ask about, without sending them to a public AI

It doesn’t stop at answers. Echo7 can also take the next step:

  • Routine workflows, automated. A question or a short form becomes an action: raising a maintenance request or defect report, drafting a handover note or report, routing a permit or request for approval, and letting the right person know.
  • Connected to the systems you already use. Echo7 is system-agnostic. It fits around the systems already on board and ashore – planned maintenance, document management, Microsoft 365 and SharePoint, and permits and safety records – without changing your existing stack. Integrations with systems on the site network stay on site. Connections to shore systems, and document updates from shore, use the link or happen in port, with your approval, and are not required for answers from your on-board documents.

Back to 03:00 in the engine room. The engineer asks Echo7 what to check first. It finds the procedure in your own manual, points him to the section, and the question never goes over the satellite link.

What it changes for IT, HSE and operations

  • Data stays on site. Manuals, logs and incident reports aren’t uploaded to a public AI service.
  • A smaller attack surface. One less outside AI service to connect to, secure and account for in your cyber risk assessment.
  • No reliance on the link. Answers keep coming when the connection drops, and AI uses none of your satellite bandwidth.
  • High availability if you need it. A resilient set-up is available for both on-premises options if a site can’t afford downtime.
  • It sits on top of what you have. Echo7 is system-agnostic and works alongside your existing systems rather than replacing them.

To be clear about what it isn’t: on-board AI is not a cyber security programme. It won’t stop ransomware on its own, and it doesn’t replace patching, segmentation or monitoring. What it does is remove one of the newest and least controlled routes for data to leave your site, and give crew a safe alternative to the tools they’re already using.

Not every site needs AI on board

Echo7 and the AI model are separate parts. You choose where each one runs:

  1. On-premises, local AI. Echo7 and the AI model both run on board, nothing leaves, and no internet is needed. This is the option for vessels, rigs and platforms.
  2. On-premises, external AI. Echo7 runs on your own hardware on board, so your documents stay on site. Only the question and the relevant extracts go to the external AI model for each answer.
  3. Echo7 Cloud. Hosted by Echo7 with an external AI model. No hardware on site and the fastest start, for teams with reliable connectivity. A dedicated cloud option is available if you prefer to host Echo7 in your own premises.

Keep the AI. Keep the data.

Every offshore team ends up asking the same three questions: can we use AI, is our data safe, and will it work when the link goes down?

Your crew will use AI either way. You can choose which tools they use and where your data ends up.

Talk to our offshore team about which deployment fits your sites.

Related posts

Unlock the power of AI with your data

Or ask our assistant a question