What is the foundation for secure use of ChatGPT?

Depending on how you use ChatGPT, you can get answers that are at least as good as experts, and it is expected to be used in various fields. On the other hand, if security measures are not taken, confidential information may be leaked, and other damage may occur. This article mainly introduces security measures when using ChatGPT in an enterprise.

Is it dangerous to access ChatGPT via the Internet?

ChatGPT is getting a lot of buzz these days. I'm sure some of you use it every day.
Normally, when you use ChatGPT, you access it over the internet. However, when security is a priority, such as in a corporate setting, can you just use it as is? You need to be careful when you use ChatGPT over the internet, because you will be exposed to a variety of risks, such as;

  • Virus infection
  • Being compromised
  • Information leakage due to negligence

Figure 1: Risks of connecting to ChatGPT over the Internet

So, you might be thinking, what can we do to mitigate the risks and use ChatGPT?
One solution is to create a secure closed network [*1] environment which will be explained in the next section.

  • [*1] An environment that is isolated from the public Internet.

An overview of secure closed network environments

I'm going to start with a bit of technical jargon.
Setting up a closed network environment means users need to reach ChatGPT without exposing it to the public internet.

To give you a concrete idea, let's look at an environment that combines Azure OpenAI [*2] and various Azure services that you can build with our templates.

  • [*2] ChatGPT service in Azure

Figure 2: Connecting to Azure OpenAI in a closed network environment

Compared to Figure 1, there are various services in between, but there are three important points.

(1) Private Endpoint for using Azure Private Link
Using Azure Private Link, you can access Azure OpenAI through an Azure managed network without having to go out to the internet. Access to Azure OpenAI will go through the Private Endpoint shown in the diagram.
(2) DNS Private Resolver
You can access the Private Endpoint's private IP by connecting with ExpressRoute or VPN from outside the closed environment. However, when accessing with the Azure OpenAI URL, the private DNS zone is not accessible, so name resolution is not possible, and the Private Endpoint's private IP cannot be obtained.
As a solution to this problem, you can perform DNS queries to the private DNS zone from the customer location by using DNS Private Resolver.
(3) Azure OpenAI Service
According to Azure's terms of service, it is explicitly stated that the input and output information of the Azure OpenAI Service will not be used for model improvement or other purposes. Additionally, the Azure OpenAI Service monitors for inappropriate use that violates the code of conduct. By submitting an opt-out request for misuse monitoring, access to input information by OpenAI operators can also be restricted. By appropriately using the service and submitting an opt-out request, you can prevent input/output data from being used for model improvement and avoid information leakage to external parties. While this differs from risks posed by internet access, it is equally important to prevent information leakage from this perspective.

Summary

  • When using ChatGPT in an enterprise, there are security considerations. One of them is access via the Internet.
  • In this article, we introduced a method for constructing a closed network environment. NTT DATA supports the quick construction of an environment by preparing a construction template.
  • In addition, we can help you start using ChatGPT with a sense of speed by fully addressing security considerations, not only for closed networks.
  • Furthermore, NTT DATA provides a wide range of support, from examining use cases for using ChatGPT to building environments and actually using ChatGPT in companies. If you want to use ChatGPT securely, please feel free to contact us! Let's create new business value together.
Masahiro Baba

Masahiro Baba

NTT DATA Group Corporation

As an IT specialist, he is involved in proposal, development, and operation of projects that utilize atypical data. He is always thinking about how to respond to changes in system requirements and data.

Yuto Yoshida

Yuto Yoshida

NTT DATA Group Corporation

After working as a field SE, he currently engages in technical support for customers as a tech lead in the field of data & intelligence. He studies the concept of a data platform using data virtualization, etc. and participates in projects constructing such platforms.

Tatsuya Sakamoto

Tatsuya Sakamoto

NTT DATA Group Corporation

He was involved in e-commerce sales and new business development at a telecommunications company. He currently engages in pre-sales in the data & intelligence field at NTT DATA.

Kantaro Narihara

Kantaro Narihara

NTT DATA Group Corporation

After being involved in development of host and peripheral systems in the financial sector and overseas projects in Singapore and Germany, he currently engages in technical support in the field of data & intelligence as an IT specialist.

Kaori Kikuchi

Kaori Kikuchi

NTT DATA Group Corporation

Started as a server-side engineer at a venture IT company, then joined NTT Data. After engaging in a variety of customer support activities as a consultant, she now mainly provides basic support for data utilization projects.

Contact

D&I Technical Department Offering Contact: dioffering-contact@kits.nttdata.co.jp