Big Law’s AI Challenge Drives New Conversation About Training

Michael Best & Friedrich blocks its lawyers and staff from using ChatGPT on the job.

The law firm made the decision to ban internal use of open-source generative AI tools after consulting some of its clients, according to Sarah Alt, Michael Best’s chief process and AI officer. The move also followed feedback from the firm’s insurers, she said.

A key question the firm needed to answer, Alt said, is could Michael Best use the power of such publicly accessible programs while making sure it’s “a safer place” for staff and clients? The answer was “no,” she said, especially given that many client companies are taking a cautious mindset about ChatGPT because of data concerns.

Other major firms are taking different a different tack, allowing limited use of the technology on the job. Some are opting to build or buy their own AI tools.

What most agree on is that lawyers and staff need to be trained on the technology. Fast.

“Developing that technical competence has become increasingly important in the legal profession,” said Jeffrey Chivers, who teaches an AI course at Yale Law School. Law firms are “error-intolerant” businesses, which means attorneys need to understand the technology that they—and their clients—are using.

Firms are trying to quickly devise comprehensive internal training regimens, even as the technology advances. They want to avoid pitfalls that range from from exposing sensitive client data to inaccurate information, like the AI-generated court brief littered with wholly fabricated court decisions that recently landed two New York lawyers in hot water.

Some firms are offering in-person seminars taught by outside specialists and video-learning presentations from law firm training platforms.

Orrick Herrington & Sutcliffe has partnered with tech training provider AltaClaro to develop a “prompt engineering” curriculum being rolled out across the firm, said Daryl Shetterly, managing director of Orrick Analytics. Prompt engineering involves structuring and refining the queries that produce AI-generated content.

The training includes instruction on how generative AI programs work, Shetterly said in an email. It also “highlights the importance of data privacy, confidentiality, checking accuracy, and other issues folks are likely to encounter.”

As firms wrestle with the new tech, several are in the process of developing policies that aim to lessen the risks generative AI tools pose.

Early this year, for example, BakerHostetler issued a firmwide directive advising its staff not to use large language models combined with any client data, said Katherine Lowry, the firm’s chief information officer. Data ingested by the public ChatGPT and similar tools isn’t kept private and could be breached, according to the firm.

Firm leaders are “deep in discussions” over a broader policy regarding chatbots, Lowry said.

Law Schools

US law schools are also working to meet student demand for up-to-date courses about the fast-evolving tech.

“Students are hungry for it, and professors are aware of it,” said Dyane O’Leary, chair of the Association of American Law Schools’ section on technology, law, and legal education.

O’Leary will be teaching a new winter intercession course—Generative AI for Lawyers—at Suffolk University Law School. Elective classes are one thing, but O’Leary says there are many students they won’t reach.

The “huge trick,” according to O’Leary: “Is this going to be integrated into the existing law school curriculum?”

Arizona State University’s law school gained attention in July when it announced that it will allow prospective students to use chatbots to help prepare their applications.

“Firms are looking for candidates comfortable and knowledgeable in this space,” ASU Law dean Stacy Leeds said in an interview.

—With assistance from Pavithra Rajesh