Getty Images/iStockphoto

ChatGPT Enterprise targets some enterprise concerns, not all

While the new product does not use customer prompts or company data in training OpenAI products, it fails to address other concerns surrounding hallucinations and copyright.

OpenAI is looking to ease some concerns about its popular generative AI product ChatGPT.

The AI vendor on Aug. 28 introduced ChatGPT Enterprise. It is generally available now to businesses. ChatGPT Enterprise differs from the consumer version of ChatGPT in enterprise-grade data and security.

The enterprise version enables organizations to configure the training of their model and how long internal data can reside in the model.

Also, ChatGPT Enterprise encrypts data and provides single sign-on authentication and the ability for enterprises to manage who has access to certain levels of data. Customer data won't be used to improve OpenAI services, and the product will serve an assistant for enterprises, the vendor said.

The latest changes, embodied in ChatGPT Enterprise, comes after OpenAI made several incremental updates to ChatGPT over the past few months.

Earlier this month, OpenAI expanded custom instruction access to free ChatGPT users, enabling them to add preferences or requirements for the model to consider when generating responses.

OpenAI also rolled out a code interpreter feature in July to ChatGPT Plus users. Earlier in the year, the Microsoft-backed vendor released ChatGPT App for iOS and Android. And in the vendor updated ChatGPT's data settings so users could turn off their chat history and choose if they want their conversations to be used to train and improve OpenAI's models.

OpenAI said ChatGPT Enterprise is aimed at helping employees learn concepts such as coding and how to analyze corporate data. 

The concerns not addressed

Despite being a positive step, ChatGPT Enterprise still needs to address some of the other concerns enterprises have about the generative AI system, according to Jim Hare, an analyst at Gartner.

One worry is that while ChatGPT Enterprise uses OpenAI's latest large language model, GPT 4, it is still unclear what training dataset is used to train the model. Moreover, enterprises concerned that the model may be trained on copyrighted material, due to some recent lawsuits against OpenAI, may be skeptical about immediately adopting the new enterprise product.

Also, the tendency of large language models (LLMs) like ChatGPT to "hallucinate" and misrepresent facts is still a concern, especially if enterprises plan on using the technology in customer-facing applications such as a customer service center or if customers are interacting with it.

"I think it's really addressing some of the risks, but there are other ones," Hare said. "I will [advise] enterprises to take a look at it, experiment with it, but don't look at this as the be-all-end-all that's going to solve all these enterprise problems."

Also unclear is the price of ChatGPT Enterprise, Hare noted. OpenAI did not disclose how much it will cost enterprises.

Opportunity for enterprises and OpenAI

Regardless of these challenges, ChatGPT Enterprise could be an opportunity for some enterprises. For example, although Microsoft has Azure OpenAI, another enterprise-grade generative AI product for those interested in OpenAI technology and ChatGPT, some enterprises might not be "Microsoft shops," Hare said.

It's really addressing some of the risks, but there are other ones.
Jim HareAnalyst, Gartner

Moreover, the enterprise version is a way for OpenAI to capitalize on some of its investments while continuing to educate people on generative AI with its consumer products.

"The consumer stuff is great for individuals," Hare said. "The enterprise is OpenAI's attempt to really monetize."

While general purpose LLMs like ChatGPT and Google Bard are still getting a lot of attention, the trend is moving towards domain-specific models, Hare continued. These models will mitigate the need for prompt engineering because they are built for specific business applications such as marketing, sales or customer support.

"Enterprises will look at what OpenAI is providing and say, 'Well, that's great.' But they'll lean more toward these providers that are building models that are more geared for their particular industry or their business," Hare said.

OpenAI said it plans to release more features to ChatGPT Enterprise, including tools for specific roles such as data analysts, marketers, and customer support.

Esther Ajao is a news writer covering artificial intelligence software and systems.

Next Steps

OpenAI's new versions of GPT target more enterprises

Dig Deeper on AI technologies

Business Analytics
Data Management