AI: For Better or For Worse

When attorneys at Seyfarth Shaw recently shared a legal brief with a client, the customer had positive feedback—for the most part. 

“One of the comments was that our firm should have made the client aware of a decision in the same court where the judgment was entered against a competitor on the same issues,” says Brett Bartlett, who practices labor and employment law at the Atlanta firm. “Our client said, ‘Why didn’t you tell us about this other case?’”

For a good reason: It didn’t exist. The client’s in-house counsel had run the brief through an AI platform, which “hallucinated”—or made up—additional cases that might be relevant. This can happen when an AI platform is asked to handle a task beyond its training.

In an effort to work faster, smarter and cheaper, more and more businesses are attempting to navigate the evolving landscape of artificial intelligence.

“AI provides a useful tool to help clients accomplish any variety of things in the workplace—so long as AI is being used to augment responsibilities rather than replace them,” says Bartlett. 

The basic technology has been around for decades. In the mid-1990s, robot pets were introduced—think Tamagotchi and Furby—and IBM’s Deep Blue computer made headlines by beating chess grandmaster Gary Kasparov. OpenAI released its first GPT (generative pretrained transformer) in 2018, demonstrating how the technology could perform language tasks like researching large amounts of data and composing text. By 2022, ChatGPT was introduced, allowing people to interact with the large language model (LLM) via chatbot. 

“All of our clients are considering how to use AI, and we’re helping them assess how they might use it,” says Bartlett.

Not surprisingly, as the use of AI increases, so do related problems and concerns. While many embrace the promise of ChatGPT, Microsoft Copilot and other AI applications to enhance productivity, learning, and even creativity, others fear its potential to rob human beings of, well, all the above. Not to mention their jobs.

Upsides

One of the main upsides of AI is its potential for saving time and money. 

Katherine Silverman, a business litigator at Berman Fink Van Horn in Atlanta, works primarily with small- to mid-size businesses, including commercial real estate owners. “If you’re someone who’s been leasing property for years, you can take all of your leases and put them into an LLM. If you need to draft a new lease and you want to know how you’ve handled repair and maintenance clauses in past leases, or you want to draft a new one that matches what you’ve historically done, that can be a great use of AI,” says Silverman.

Technology transactions attorney David Schulman, with Greenberg Traurig in Atlanta, sees a similar time-saving benefit for engineering firms he represents, some of which have been in business more than 50 years. 

“There’s a trove of data and information and know-how that exists in historic material,” he says. “Through data mining and running them through AI models, you can improve the efficiency and the quality of your deliverables now that you might never have had access to because information was sitting in drawers or in archives.” Of course, those overlooked resources must first be digitized.

Schulman’s clients also include video game-development studios. “AI has allowed indie video game studios to become significantly more nimble and save a significant amount of money,” he says. He’s seen his smaller clients utilize AI to create products that look as good as ones created by AAA companies at a fraction of the cost.

Bartlett sees similar possibilities for human resources departments at the companies he represents in the retail, finance, hospitality and hotel sectors. “When you have a business that has locations across the country, and you have to consider whether to interview hundreds of job applicants, AI certainly seems like salvation, to some extent,” he says of the tool’s ability to review applications and applicants according to preselected preferences. “Most of the hiring managers across big companies have day jobs beyond just hiring. If they can use AI to make their lives easier, they might do so.”

Downsides

Such savings can disappear if someone inadvertently uses AI-generated data that is copyrighted, inaccurate or fictional. 

Schulman tells his clients that AI output needs to be vetted. “Assume the situation is like a junior member of your team handing you the output,” he says. “Examine it with the right level of scrutiny to make sure that the output is appropriate for your needs and not biased in ways that are potentially damaging to your business and to your customers.”

Attorneys caution that AI efficiency comes with potential hazards—like reputational risk, legal liability, potential bias in hiring decisions, and the possibility that a company’s own proprietary information could be compromised.

Bartlett’s clients often ask about potential biases when using AI for hiring decisions. He and his peers are paying attention to a California-based lawsuit involving AI human resources platform Workday, used by more than 11,000 international companies. In Mobley v. Workday, Derek Mobley, a Black man in his 40s, alleged that Workday engaged in age, race, and disability discrimination after he was rejected for more than 100 jobs by companies that use Workday’s screening features. Workday filed a motion to dismiss Mobley’s suit, but Judge Rita F. Lin denied the motion and is allowing the case to proceed as a collective action lawsuit.

“The common view among most of us employment lawyers is: To the extent a human can engage in discriminatory decision-making, so can AI,” says Bartlett. He also advises clients on matters of employee classification. As workers use AI to complete tasks that eliminate some of their duties, managers are questioning how this use might affect exemption from overtime pay requirements. 

As for the concern about proprietary information, Silverman urges businesses to examine how their workers use AI. “If you’re letting your employees use personal phones or personal computers to do company business, if they’re working remotely, what ability does the company have to control and restrict the employee’s reliance on and use of artificial intelligence services?” asks Silverman.

Silverman advises clients to educate employees about the appropriate and legal use of technology like ChatGPT. This can help avoid problems like inadvertently sharing the business’ proprietary information in a data search. She also urges putting policies in place so the company can defend itself if an employee does engage in improper behavior while using AI. Finally, she recommends that clients use a subscription or enterprise level of AI whose terms of use have been reviewed by the company’s IT department or service to make sure they offer sufficient protections for the business. 

“Often the paid version of something is going to be more secure than the free versions,” says Silverman.

Intellectual Property Concerns

Schulman, whose video game clients create franchises that can bring in millions, is especially attuned to the potential compromise of intellectual property rights when AI gets involved. “You don’t get the same intellectual property ownership rights over a product created with AI that you would if it was human-generated,” he explains. “The AI models get their data by scraping the internet, and it is possible that the input that it used to create its algorithm is infringing. There is also a risk that the output for the AI model infringes the copyright of the author whose information was used to train the AI model.” 

As a result, when video game clients ask if they should use AI, Schulman advises them to consider the possible compromises and risks. “The main characters in the game, the major undertakings in the game, should be human-generated so they’re copyrightable and protectable,” he says. “The background, the levels—things that are not necessarily unique to your game—you can use AI, but understand you don’t have a copyright in that.”

Both Schulman and Silverman point out the risks to clients’ proprietary information when using AI. “They should consider a license with the AI model provider that will [restrict] access to the data information that they’re feeding into the AI model,” says Schulman. He also recommends clients assemble a committee to create an AI usage policy, just as they would with HR or internet policies.

“Anything where you have sensitive data that’s subject to patent copyright, you need to be careful,” adds Silverman. In the real estate sector, that could include trade secrets like client lists or ways of valuing properties. This should be clarified for workers, she notes. “What employees understand to be a trade secret versus what lawyers understand a trade secret to be are often very different things.” 

Added to that, says Schulman, is the fact that many AI models are trained on North American and European data, which suggests inherent bias. 

The Long View

“As with all new technologies, there’s a process of examining the capabilities of AI,” says Schulman. “The capabilities are also changing and evolving very quickly, so there’s a constant need to examine them and to ask: What is it that a particular tool can bring to a business?” The way a business uses AI needs to evolve as the technology becomes more sophisticated. 

Bartlett recommends involving legal counsel early on. “My advice to employers who are considering the use of AI is that they should get their in-house or external counsel involved on the front end rather than a kind of check-the-box, last-minute thing,” he says. “These are complicated platforms. They should be tested to ensure that they’re doing what is represented they’ll be doing.”

He suggests taking the long view on this relatively new technology. Bartlett’s own firm—much like others, he suspects—has spent a considerable amount of time discussing how best to use AI. He compares it to using email, LexisNexis or a calendar function. “It’s something that we’re able to use to augment what we’re doing,” he says. “It’s a tool.”

“Most of my clients are in the discovery and learning phase,” says Silverman. “It can be a big commitment. I encourage clients to be aware of what’s out there and to investigate.”

Taking AI to Court

Thomson Reuters v. Ross Intelligence

In perhaps the earliest AI-related case, media giant Thomson Reuters sued AI startup Ross Intelligence in 2020, claiming the new competitor had violated copyright law by using material from the Thomson Reuters’ Westlaw legal research platform to train the startup’s AI search tool. The court ruled in February 2025 that Ross’ actions were not fair use.

Advance Local Media LLC v. Cohere Inc.

More than a dozen news corporations, including Atlantic Monthly and Conde Nast, sued Cohere in early 2025, claiming the Canadian AI startup committed copyright infringement by using thousands of their articles without permission. In addition, the suit claims that Cohere hallucinates inaccurate articles and attributes them to the news companies’ writers.

State of Texas v. Meta Platforms, Inc., f/k/a Facebook, Inc.

Meta (formerly Facebook) agreed to a $1.4 billion settlement with the state of Texas in July 2024 over allegations that the company collected facial recognition data from customers without their permission.

Lawsuits against various insurance companies

Health insurers including United Health, Cigna and Humana are being sued in class actions claiming they used AI algorithms to deny care for customers without consideration of individual situations. The suit against United Health, Estate of Gene B. Lokken et al. v. UnitedHealth Group, Inc., for example, accuses the company of using nH Predict to deny extended care for the deceased Medicare Advantage patient.