On Tuesday, Microsoft and OpenAI were named as defendants in yet another class action lawsuit for allegedly breaking several privacy laws.
This is the second class action lawsuit that the two companies have faced since the beginning of the year, which was marked by the start of their multi-billion dollar partnership.
A filing registered with a U.S. Court accused the two companies of using web scraping techniques for stealing private information of millions of internet users, including children, without their informed consent or knowledge, for training ChatGPT and other AI models.
Earlier this year, OpenAI was also accused of leaking intellectual property of ChatGPT users, leading a number of companies to ban the use of the bot in the workplace.
Soon after, Microsoft announced Azure ChatGPT, an enterprise solution, which it touted as “your private ChatGPT,” and at the same time, acknowledged the shortcomings of OpenAI’s ChatGPT in a now deleted post.
This came while OpenAI was preparing to release its own enterprise version of ChatGPT.
Rumours of a rift between the two companies continued to swirl as OpenAI’s chief executive Sam Altman went on to clarify on X that OpenAI doesn’t use API-submitted data to train or improve models unless a user explicitly opts-in.
Azure ChatGPT has now been deleted from GitHub’s repository and Microsoft denied the existence of “a product known as Azure ChatGPT,” in an interview with Analytics India Mag. But that was enough to reveal a complicated relationship between the two companies.
Is there drama?
Microsoft’s collaboration with OpenAI was the perfect tech give and take. OpenAI would build the foundation models that would power Microsoft’s services, and Microsoft would undertake the cloud costs on Azure.
But OpenAI’s privacy gaffes, colossal computing costs, and mounting losses has Microsoft edging away slowly, an Analytics India Mag report suggested.
“Microsoft is concerned about the current negative perception of OpenAI’s data security for enterprises,” the report noted. “Through the introduction of Azure ChatGPT, they are striving to rebuild trust among enterprises and attract more customers. They don’t want to associate themselves with OpenAI’s name on Azure, that’s why they have put Azure’s name at forefront.”
But there’s no tension, Forrester senior analyst Rowan Curran argued, but rather just an attitude in the market right now, whereby people have a negative perception of OpenAI while they view Microsoft as a trusted vendor.
He added that the perception that OpenAI is “an insecure platform that’s going to leak all your data is just not true.” Whether you use OpenAI APIs on Azure or on OpenAI directly, you are going to get the same type of data sharing agreement that you would expect from using a free service, he said, stressing that the enterprise stuff will take time to get set up.
Though Microsoft deleted Azure ChatGPT, its plans to sell a new version of data firm Databricks’ software to help customers create AI apps for their businesses, would, however, again place it in direct competition against OpenAI’s offerings, explained GlobalData technology analyst Beatriz Valle. So would Microsoft’s upcoming AI-driven collaborative software, Copilot, which will compete with ChatGPT Premium.
Microsoft’s recent association with Meta’s Llama 2 also appears to have strained OpenAI’s relationship, creating a competitor for its closed source models.
But for Curran, that means that Microsoft just recognized that open source is going to have a very significant role to play in the large language model ecosystem and to address the needs of all its customers. Llama’s family of models, he added, is one of the best options available to companies – a commercially usable model that they can run on smaller footprint and more local hardware.
Further, Valle noted that OpenAI is in “a vulnerable position due to revenue imbalances resulting from the massive spend of its cloud computing usage”, but given it is still early days in the generative AI market, it will take time to gauge the level of adoption of offerings such as ChatGPT Enterprise.
Regardless, OpenAI’s ability to turn its business into a profitable model, Curran added, does not matter as much, in the short-term, as its success with being a technology enabler to Microsoft.
“It’s important to recognize that part of what Microsoft was investing in OpenAI for was basically to have the technology fed into Azure, so that Microsoft didn’t necessarily have to build that technology in house and did not necessarily have to inherit all of the organizational and cultural baggage that is present in a larger company,” said Curran. “So I think just focusing on whether OpenAI is profitable disregards the fact that they’re on a path forward right now, in being the technology enabler to their key partner right now, which is Microsoft.”