Boutique law firm builds its practice around AI risk

A Washington, D.C., law firm has built its practice around artificial intelligence, betting that companies using machine-learning tools will need more legal and technical scrutiny. Fortune reported that bnh.ai is focusing on risks tied to privacy, security and fairness as lawmakers signal more rules for AI systems.

The firm describes itself as a boutique practice for both lawyers and technologists, according to Fortune. Its founders argue that AI legal work requires more than a conventional understanding of software because machine-learning systems can change as they absorb new data and user behavior.

Fortune reported that companies need to know what data trained their systems, how those systems may shift over time and how attackers could tamper with them. Researchers continue to identify new ways to interfere with machine-learning software, the publication said, adding complexity for businesses trying to assess legal exposure.

Andrew Burt, managing partner of bnh.ai, told Fortune that one obstacle is the communication gap between attorneys and data scientists. He said lawyers can find technical discussions intimidating, while technologists may feel the same way when lawyers rely on specialized legal language.

Burt said the “future of technology is dependent on those meetings” between the two groups, according to Fortune. In his view, lawyers must understand enough about AI systems to explain legal risks in a practical way and help data scientists address them.

That work differs from oversight of many traditional software products, Burt told Fortune. Machine-learning systems require continuing review because their behavior can evolve, creating risks after deployment.

Fortune reported that Burt expects early conversations between lawyers and technical teams to be awkward. He said technologists may resist legal involvement and may not want to spend time on uncertain problems without clear answers.

Burt also described a cultural split between the professions, citing a point from his co-founder Patrick Hall. According to Burt, Hall has said lawyers may see value in a meeting devoted to discussing legal issues, while data scientists may feel time was wasted if the meeting produces no code.

Despite that divide, Burt told Fortune the groups can work together once they establish a shared frame of reference. The firm’s work can include questions such as how to divide populations within datasets under fairness rules and when to retrain a machine-learning model with more relevant data.

Burt said companies should address AI risk before a failure draws regulatory attention, Fortune reported. He said the business leaders most prepared for AI will be those who both invest in the technology and pay close attention to its legal consequences.

This story draws on original reporting from Fortune.