OpenAI has launched a version of its popular chatbot ChatGPT for enterprise. Unlike the free and pro versions, it includes data security in transit and inputted data will not be used to train the company’s AI models. The AI lab says it also enables customers to access its GPT-4 model with no usage caps.
The enterprise version of the popular chatbot has been promised by OpenAI for some time but the company says it had to work out security and features for big business. The new release will also include high-speed performance and access to data analysis tools.
Since its launch in November 2022, ChatGPT has gone on to become one of the fastest-growing consumer products and inspired large companies like Google, Microsoft, IBM and Salesforce to launch or expand their own generative AI offerings. There is a solid business reason for this change as, according to Bloomberg Intelligence, the generative AI market will be worth $1.3trn by 2032.
IBM has already launched its own enterprise AI offering in the form of watsonx, which includes a code generator and other generative AI services. Microsoft also has an enterprise version of Copilot, which is built on GPT-4 from OpenAI.
OpenAI may face a problem convincing customers its systems are safe to use. Companies like Samsung, Apple and Goldman Sachs have stopped employees from accessing the public version of ChatGPT, and others are exploring custom-built alternatives.
Despite this, nearly half of workers responding to a recent survey by professional networking platform Fishbowl say they’ve used ChatGPT to complete tasks at work. OpenAI says it has seen significant demand from within organisations for its tools, and that teams in over 80% of Fortune 500 companies have registered for ChatGPT accounts. This is based on accounts linked to a corporate email address, so may not be official or sanctioned.
Partner companies named as early adopters of ChatGPT Enterprise including Canva, Estee Lauder and PwC.
ChatGPT Enterprise: data security and encryption
To try and combat the data security issue, OpenAI has implemented AES 256 encryption for data inputted and generated by ChatGPT Enterprise while at rest and TLS 1.2+ for data in transit. It is also compliant with SOC 2, a voluntary standard designed by the American Institute of Certified Public Accountants that specifies a high level of customer data security and management.
“You own and control your business data in ChatGPT Enterprise,” the company said in a statement on the release of the new enterprise version. ”We do not train on your business data or conversations, and our models don’t learn from your usage.” It is possible to turn training off in the public version of ChatGPT, but it is switched on by default.
One of the biggest differences, aside from unlimited GPT-4 access and improved performance, is the significantly larger context window. This refers to the AI’s memory, and the longer the context the more data it can store before you have to remind it what you have previously asked. The public version of ChatGPT has a limit of approximately 8,000 tokens, or about 6,000 English words, whereas the enterprise version has a 32,000 limit, or 25,000 English words.
This longer context will allow for complex and lengthy reports to be shared with the chatbot for analysis, OpenAI says. It equates to about 50 standard A4 pages of text, and although it currently can’t handle images or graphs, the company says this feature will be added in future.
Other upcoming enterprise updates will include the ability to customise and extend the base ChatGPT model on specific company data. This could include connecting existing applications such as those from Salesforce. The company is also extending and enhancing its data analysis tool, previously known as code interpreter to allow tailored content for specific roles including marketing, customer support and sales.
Pricing details haven’t been released for the enterprise version but according to OpenAI, it is available now. A small business version is also in the works.