• Home
  • News
  • Promoted
  • Fujitsu: removing data biases with on-premises generative AI
igamingnext photo
Nick McDonald, account director for UK&I Platform at Fujitsu, outlines the reasons why iGaming organisations should be shifting their focus towards private GPT models to streamline their day-to-day operations.

The meteoric rise of artificial intelligence technology has been undeniable within the iGaming industry. From automation through to risk management, businesses within our sector are looking for new ways to incorporate AI within their daily operations.

Particularly over the last few months, many iGaming companies have turned their attention towards generative AI models, more specifically how these models can reshape the future of their organisation.

But what has become clear during that time is that relying on public Large Language Model (LLM) AIs – such as ChatGPT, Bing, Claude and Bard – on public cloud platforms could bring with it its own unique set of risks.

In the majority of cases, experiences in the public cloud around feature richness and velocity of innovation have been second to none.

However, more companies are beginning to address their concerns around accuracy, bias, intellectual property leakage, data privacy, emerging regulation and compliance, legal risks, and exponential technology costs.

Public LLMs, of course, do have their upsides; but they are often reliant on finite training data sets which, when prompted to generate content, may provide answers which are too generic, inaccurate, or even amplifying data biases.

Beyond the risks of using public generative AIs, public cloud platforms can also present a whole host of security risks such as data privacy breaches, IP leakage, and operational risks. One way to overcome some of those hurdles is by exploring the use of on-premises generative AI systems, or Private GPT models.

Fujitsu has been making waves in the world of Private GPT, having showcased our flagship model at the World AI Cannes Festival earlier this year. We have been working alongside our partners & customers to develop a generative AI model that can be tailored specifically to that organisation, with all data stored on-premises.

We make sure that each Private GPT model is tailored to each individual company. The best way to explain this is by looking at a case study – let’s take the legal industry as an example.

As with any business, legal companies don’t want their information being uploaded to a public form, such as Chat GPT. They want, and more importantly need, that information to remain confidential.

Our model, however, allows you to upload your own data to the private system within your data centre, that sits on the hardware stack supplied by Fujitsu. What this means is that the data sources can be much more targeted to that individual organisation and can include pools that would not be available to a public AI—for example, information held in Teams or on an intranet or extranet. 

Data about products and services can also be updated as needed; our clients can also create models that are ring-fenced to the needs of specific business units or departments.

The great thing is you can then query that data yourself. If you uploaded a legal document that was several hundred pages, for example, you could ask the Private GPT model a question and it would source the information from that document. 

When you query the data, it will provide you with a link to the page reference so that you know exactly where the information has been sourced, allowing you to understand the context to the answer too. That’s much more accurate than public GPT models; it also enables you to remove any biases / wrong answers too. 

Some companies may have their servers based in their own office environment; others may have them based in a colocation data centre. But since the system is hosted entirely on your own network, it is entirely private and only available to those working within your company. Even to the point where you don’t need internet connectivity.

The models are already trained in the background, which helps make the product ready-to-use for customers. It is extremely easy to use and has the potential to be incredibly useful within this industry.

As more industries begin to place the spotlight on security and data privacy, it’s likely that private GPT models will become more commonplace. Our Private GPT model is a ready-to-go solution and are ready to roll it out to more of our customers in 2024 and beyond


Nick McDonald is an accomplished account director with seven years of experience in the iGaming industry heading up the iGaming team within Fujitsu. Nick has worked with some of the largest platform providers in the space and has a proven track record of delivering exceptional results for clients.

As an expert in the field focused on Data and Hybrid cloud solutions, Nick McDonald is well-versed in the latest industry trends and best practices with a passion for delivering flexible and reliable technology solutions globally, and has a keen eye for identifying opportunities to help drive growth and revenue in the iGaming space through technology.

Similar posts