IBM could be a trend-setter by offering customer protection on its AI offerings

By Therese Poletti

IBM and Microsoft are hoping to make customers less nervous about using AI.

Last week, IBM (IBM) announced an indemnification pledge aimed at helping its customers overcome their fears about adopting artificial intelligence into their corporate systems, and it could be setting a trend that others follow.

Since the massive surge of interest in AI adoption, there have been a wide range of concerns from both companies and consumers about using AI, from accuracy to copyright infringement to security risks. OpenAI, the creator of ChatGPT, has been hit with lawsuits by authors contending that its large language models have been trained in part with unlicensed copyrighted content. Microsoft Corp.'s (MSFT) GitHub and OpenAI have been sued over AI-enabled software piracy.

"The biggest challenge in generative AI is the cost aspect," Dinesh Nirmal, senior vice president of software at IBM, told MarketWatch in an interview. "You need a lot of GPUs (graphics processing units) so a lot of customers are asking for a pre- trained model," so they don't have to go and train their own models. "If I have a pre- trained model like IBM, then I can take that and use my data and fine-tune it. But the challenge is, I don't know what kind of data has been used."

So as part of IBM's recent launch of a new set of AI software models for enterprise customers, the company is giving customers intellectual-property protections for the models developed by IBM. The hope is that it will make it easier to bring generative AI into business applications and workflows, and get customers interested in its Watsonx Granite series.

"We can tell you the data that have been used, and the parameters used for training, all those things we stand by," Nirmal said. "This is all work done by IBM Research, all the training of the data has been used by IBM Research on IBM systems. We have an infrastructure in place, using the data that we are comfortable with, that's why we can indemnify it." Nirmal added that IBM has published the sources of its data in a white paper that customers can review as well, and that no one else has gone so far as to publish its data sources.

IBM has been working in AI for decades. It generated huge consumer interest in its research when its Deep Blue supercomputer beat chess master Gary Kasparov at one game in 1996, becoming the first machine to win against a grand master (Kasparov won the overall match). Then in 2011, IBM followed up with Watson, which beat human "Jeopardy" champions. Even so, it has had a difficult time capitalizing on that research and on its Watson system. After forming a Watson business group in 2014, IBM eventually sold off Watson's health assets in 2022.

So offering some guarantees for jittery customers who want to try the latest, most advanced iteration of AI appears to be a savvy move.

"I think right now it is a differentiator for IBM," said Pat Moorhead, head of Moore Insights & Strategy. "Healthcare, finance, insurance, these are heavily regulated industries and risk-averse from the start...I just can't imagine them going all in on a model where there aren't some guarantees...This takes one off the table for businesses."

IBM declined to comment on how much this effort is costing, or any potential costs going forward, but described the AI opportunity as "boundless."

"The only thing that was limiting enterprises was, 'What models do I use? How do I get a legally trusted model within my enterprise that I don't have to worry about?' Now you have a model that is certified by IBM," Nirmal said.

Customers will then fine-tune IBM's Granite with their own proprietary data, but IBM noted that it is only responsible for the data models it supplies to customers.

So far, Moorhead said, IBM appears to have the most comprehensive indemnification of its own data for business, but added that Microsoft is also offering protections for its AI-powered Copilots. Microsoft is offering protection for the Copilots it offers in Microsoft 365 suite, its Bing search engine and its GitHub developer tools.

"We will assume responsibility for the potential legal risks involved," Microsoft Vice Chair Brad Smith and Chief Legal Counsel Hossein Nowbar said in a recent blog post. "If a third party sues a commercial customer for copyright infringement for using Microsoft's Copilots or the output they generate, we will defend the customer and pay the amount of any adverse judgments or settlements that result from the lawsuit, as long as the customer used the guardrails and content filters we have built into our products."

Daniel Newman, chief executive of Futurum Research, said companies are moving so fast with AI development that customers need some sort of assurances. "At the speed of this innovation, this will mitigate the potential downside and risk," he said.

At the same time, these indemnifications are also further evidence that the AI field is still the new Wild West, and that both companies and consumers alike still need to proceed with caution.

-Therese Poletti

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

10-04-23 0800ET

Copyright (c) 2023 Dow Jones & Company, Inc.