In this Law.com article, Rachel Miller explains how artificial intelligence has made vendor contract negotiations more complex, time-consuming, and legally demanding than ever before. Between the rapid pace of AI development, the absence of standardized contract language, and a lack of government regulation, attorneys are reporting spending more time negotiating individual vendor agreements, not less.
Rachel cautioned against overcorrecting in contracts, pointing out that overly broad AI addenda create their own complications, often setting expectations that vendors cannot meet.
“I’ve seen these multiple-page AI addendums that just sweep everything in because everyone’s anxious, and a lot of times the vendor can’t necessarily comply with everything the client wants,” Rachel said.
“I was just joking with a client that the whole point of AI is supposed to be making our life simpler,” Rachel added. “People said, ‘Oh, lawyers are going to be out of a job, and AI is going to take over.’ Yeah, right now my job is more complicated. I have more to do. I have more to write, more to figure out, more to counsel on.”
Please note that the original article is paywalled.
While artificial intelligence tools can help speed up some tasks for lawyers, the technology and the security implications behind it have sometimes added more wrinkles and time to drafting and negotiating contracts for clients.
Some attorneys report even higher billable hours from what was previously standard contract drafting and negotiating because of clients who have security and data concerns from a vendor’s use of Al.
Jason Epstein, co-head of Nelson Mullins Riley & Scarborough’s technology industry practice group, said that “demand has gone up exponentially” for helping client businesses navigate their vendor contracts, ever since Al tools came to prominence. Epstein and other attorneys told Law.com in interviews that the technology’s uncertain future, far-reaching security implications and lack of regulation are making every vendor contract negotiation take longer.
Some of the issues stem from the lack of any standard boilerplate explanation for how parties can use each other’s data with Al tools. That lack of standard expectations, combined with a dearth of government regulation, has attorneys committing more hours to ironing out the details, not only for each of their clients, but for each of the vendors those individual clients may work with.
Part of the time sink is when companies try to put several restrictions on how their vendors can use their information with Al, according to Kelley Drye & Warren special counsel Alex Schneider. Schneider said that data processing agreements-a normal part of vendor contracts for any sort of technology services-are increasingly being written with a focus on Al as companies seek to protect their proprietary data.
Unlike with privacy regulation, “we don’t have statutorily mandated contract terms [on Al], so you are getting more of a mix,” Schneider said.
Many contracts for new and existing vendors who start providing Al services are “not complete” when it comes to considering all of the potential impacts the technology can have on a business process and a business-vendor relationship, Epstein said. He said “vigilance” is required by attorneys and clients alike to look out for potential holes in contract terms as they navigate the current wild west of artificial intelligence.
Norms in the contract space typically change in a very slow, methodical fashion in response to regulatory shifts or market trends. That is not the case now, though, said Daron Goldstein, U.S. head of data innovation, privacy and cybersecurity at Withers.
Particularly when working with a vendor who focuses on artificial intelligence, companies and their attorneys are still trying to find a happy medium on just how much a client business can restrict their vendors’ use of AI to avoid data leaks or hallucination-related errors, Goldstein said.
“Most companies, businesses, processes and obviously the attorneys, are still figuring out how all of these things work,” Goldstein said. “It requires a lot more negotiation on a lot of these terms because we sort of haven’t figured out as a business world where the middle ground is (for) the need to use the data versus protect the data versus learn from it and all of those factors that balance.”
Contrary to what many expected, Al has only made attorneys’ work on client vendor contracts more essential, says Rachel Miller, senior legal director at mid-size technology and data law firm ZwillGen.
“I was just joking with a client that the whole point of Al is supposed to be making our life simpler,” Miller said. “People said, ‘Oh, lawyers are going to be out of a job, and Al is going to take over.’ Yeah, right now my job is more complicated. I have more to do. I have more to go to in, more to write, more to figure out, more to counsel on.”
Miller noted that while potential openings in contract language as to Al use can pose a threat to a business, trying to paint over those holes with overly broad language is its own problem for both vendors and the businesses hiring them.
“I’ve seen these multiple-page Al addendums that just sweep everything in because everyone’s anxious, and a lot of times the vendor can’t necessarily comply with everything the client wants,” Miller said.
Security Concerns
One of the primary concerns attorneys are navigating for their client businesses is how to prevent vendors from using proprietary data to train their artificial intelligence tools, including chatbots.
It isn’t enough to stop one vendor from training its tools on client data, though, Schneider said in an interview.
“Say they will have a contract in place that says [proprietary client data] cannot be used to train the model. What about if some third-party vendor or subcontractor is using data?” said Schneider, the Kelley Drye lawyer. “You also have to address those issues. And really, no third party should be using that information to train any model. So more broad language, I think, makes sense there.”
Goldstein, at Withers, similarly said client businesses want broad bars on the training of large language models by vendors.
For Al vendors specifically, though, he said there is a frustration that other vendors do not feel during contract negotiations, given the nature of their work. Unless one feeds large language models data, the models cannot function.
Like it or not, though, Goldstein said that many Al vendors are having to adapt to the restrictions imposed by the businesses that hire them for fear of losing revenue and missing out on at least the portion of a client’s data that they will be permitted to use for training, if there is one.
“There is a tension between the needs of businesses to keep their information confidential and the need of the large language models to get more good data,” Goldstein said. “For the most part, commercial contracts and vendor agreements say the vendor will not train their model on the data. I think that’s become the norm.”
The Unknown Valley
The signing of a vendor contract is not necessarily the end of the negotiation as far as Al concerns. Given the pace at which large language models are updated, some attorneys have found that solid contract language on Monday can become totally out of date by Tuesday.
Epstein, of Nelson Mullins, said that the constant updates to Al tools and their increasing presence in various sectors make it virtually impossible to fully predict what language may become necessary for a client’s future with a vendor.
“We all know where artificial intelligence is promised to go, and oftentimes you’re left with grasping to future proof the document to address the new issues,” Epstein said. “Once you get a deal done, very often the business does not go back to the lawyers to address the next model. That might, you know, expand the number of issues to address.”
Overall, Akerman chief executive officer Scott Myers said, AI tech “motivates clients to engage with law firms more to handle relatively new and novel risks” to a large extent.
Myers added that on top of the actual developments in AI, attorneys also need to prepare for how public perception of the technology-unfounded or not-may impact future contract negotiations between businesses and vendors.
“More and more people are starting to believe that AI is conscious and is entitled to welfare rights,” Myers said. “That matters, because whether you agree with it or not really isn’t the point if enough regulators start to believe it.”
Reprinted with permission from the June 2026 issue of New York Law Journal. © 2026 ALM Media Properties, LLC. Further duplication without permission is prohibited. All rights reserved.