Considerations for AI Adoption

As we bring AI into our businesses, we need to do it in a way that reflects our values, using it securely, responsibly, and sustainably in our industry, confidentiality is foundational.

We handle ultra-high-net-worth UNHW clients, sensitive crew data, and commercially valuable business information.

How we engage with AI, where data goes and how it's stored, are critical decisions.

Fear of data risk often holds businesses back, but we can manage those risks and still take advantage of AI's value.

AI is already used in some of the world's most secure environments, including military and government.

So let's take a look at a practical spectrum of ways to engage with AI while keeping privacy and data security in mind.

From level one, free tools like ChatGPT to level six, fully isolated, top secret systems.

For most of our needs, the green zone is where security, practicality, and cost meet.

Buying AI tools gives you fast access to powerful capabilities, but it's important to understand the levels of control and data protection they offer.

Level one free public tools like ChatGPT should never be used for confidential data.

They often train on your inputs, which can end up in public responses.

Level two, paid tools like ChatGPT Pro or Business offer better data handling, but they're still public platforms.

Best for low risk use.

Professional Software-As-A-Service SAAS platforms are purpose built with clear data policies and certifications, making them safer for business use.

Ensure that you check the platforms user documentation to fully understand how your data is managed and whether or not the underlying tools train on your data.

Level three, enterprise AI tools inside your infrastructure offer stronger data control, but may not currently match the performance of leading open source models.

An example of this could be like using Microsoft copilot inside your Word program.

Building AI systems, whether in the cloud or on-premise, gives you greater control over how data is handled alongside more flexibility to tailor solutions to your needs.

Level four, private cloud AI uses platforms like AWS giving you strong security without owning the infrastructure.

Level five fully on-premise AI runs on your own service keeping data entirely in-house, but with high costs and complexity.

And level six fully air gapped AI used for highly classified environments would be wonderful to have, but it's overkill for most business needs and comes with extraordinary cost and complexity.

When we look at cost and capability trade-offs, most superyacht businesses will find the sweet spot between level 2.5 software as a service tools and level four private AI cloud systems.

These levels give you powerful AI while keeping data protected without the need for heavy infrastructure for highly sensitive data.

You can also use a sanitise, interact, and re-identify approach.

Using this approach, you strip out sensitive details locally.

Use the AI model, then reinsert the data back inside your secure environment.

This lets you safely access AI capabilities even for high risk tasks while keeping control of the most sensitive information.

Because these models are trained on the entirety of human history.

They are by nature shaped by our past.

And our past, as we know, hasn't always been fair.

History is often written by the winners, and it carries deep rooted biases across gender, race, class, and geography.

Which means that AI systems, no matter how powerful, will inevitably carry forward some of those same biases unless we deliberately intervene.

AI can't genuinely reason, and it lacks human empathy.

We must regularly audit outputs, monitor for bias, and ensure that important decisions, especially those that affect people's lives, are made with human oversight.

The principle of human-in-the-loop is essential for transparency and accountability.

For most businesses, you'll be using AI models, not building them.

As users, you need to be mindful that you are always responsible for the outputs and what happens as a result of them.

Given our industry's commitment to sustainability, it would be negligent not to raise the environmental impact of AI.

AI's energy use is rising fast.

By 2030, data centers could consume 4.5% of global electricity, the same as multiple nuclear plants.

Big tech is acting.

Microsoft, Google, and AWS have set aggressive targets, carbon negative water positive, zero waste by 2030.

Google is also taking steps to bring nuclear energy into the AI supply chain.

At a business level, you can reduce impact by using smaller models for day-to-day tasks, right-sizing, computing power and batching processes where possible keeping prompts focused and workflows efficient, also cuts unnecessary compute and cost.

Importantly, AI can drive sustainability too.

It can help monitor, predict, and optimise operations, cutting, resource use, and emissions.

Rossinavi SeaWolf X Catamaran launched May last year is a great example where AI reduces fuel use and emissions across the vessel's lifecycle.

Reid Hoffman, co-founder of LinkedIn and a leading voice in responsible AI suggests that AI can be a powerful tool for enabling smarter, cleaner energy use globally.

I see the current movement in AI being the driver that pushes our society into clean energy adoption, fueled by the movement of big tech.