AI gebruiken zonder risico: Zo bouwen wij aan security-first AI-oplossingen • Enlite

That the use of AI agents can lead to data breaches has become painfully obvious in recent years. Fortunately, many companies now have policies in place for the use of AI chatbots, which should ensure that the number of incidents is limited. These policies result in limiting the use of chatbots. Summarizing minutes of a meeting, for example, is no longer possible because it could contain sensitive information. Customers are therefore increasingly approaching us to ask whether a secure AI chatbot can be developed. In this article we describe, how we, in cooperation with our partner Warpnet, building security first AI solutions. 

How do data breaches caused by AI occur? 

When company employees accidentally share personal data with an AI chatbot like ChatGPT, this data ends up in the databases of AI companies. Niels Gnodde, CTO at Enlite explains how this can lead to serious data breaches: “AI solutions like ChatGPT are hosted in the United States. There, very different information security rules apply than in Europe. There is no AVG or similar law in the United States, which means that a company whose data is shared with ChatGPT no longer has control over what happens to that data. 

Situations like this can lead to a data breach, unexpected disclosure to other users or access by U.S. authorities through the Cloud Act. The lack of a valid processing agreement and violation of AVG breach rules make it a serious problem. ” 

The benefits of AI chatbots 

Using AI chatbots can save a lot of time in various situations. Earlier in the article, we mentioned summarizing the minutes of a meeting. In addition, an AI chatbot can be used as a writing assistant. However, it can also go beyond that. When an AI model is aware of all the processes within a company, it can help employees complete certain steps faster. For example, consider a support company. When a ticket comes in, a support employee can use the right AI model to go through the first steps. Thus, AI can ensure that less time is lost in figuring out the cause of the problem, allowing the employee to solve the problem in a shorter amount of time. With this type of solution, it is extra important that privacy can be ensured. 

Our solutions 

From customers we are now also hearing more and more requests to build security first AI solutions. Niels: “Many companies are asking for an AI assistant for different purposes. We have built an AI model based on this demand. In doing so, we make a combination between language models and knowledge bases from which the AI model can extract all the information it needs as an assistant. We ensure privacy by shielding data from the knowledge base from unauthorized users. In addition, it is essential for privacy that the application is hosted locally.” 

Enlite builds the software and Warpnet picks up the hosting. Roald Nefs, CTO at Warpnet explains, “Security is not an afterthought for us, it is at the core of how we develop. In our partnership with Enlite, we bring in-depth platform engineering expertise and apply security by design: from secure infrastructure to local AI hosting that fully meets the customer's compliance requirements. Enlite then builds an application around that, on a foundation that is set up right from day one.” 

Develop from expertise 

Enlite and Warpnet previously developed a number of applications in which security plays a key role, Secure Dropzone and Logspect being examples. Secure Dropzone is a solution that allows organizations to share documents without risk. It is designed with multiple layers of security and files are stored on European servers, ensuring AVG compliance. Developed in collaboration with the Martini Hospital, Logspect ensures that unauthorized access to patient records can be controlled. With this, the privacy of (medical) data can be guaranteed.  

From this expertise, Warpnet and Enlite continue to develop applications such as the AI model. Niels: “The goal is to bring more AI solutions to the market that are fully customizable to the needs of our customers. Security is always the starting point in this.” 

Also using security first AI solutions from Enlite? Get in touch!  

Or would you like to help build our latest AI solutions? Check out this position!