AI is creating a professional liability exposure that’s difficult to separate from ordinary human error: practitioners may rely on generated work without catching fabricated or missing information, and firms may not know where AI entered the process. Insurers generally treat resulting negligence under existing errors and omissions policies unless the policy specifically excludes AI, but that position could change as claims develop.
“A real exposure for errors and omissions” can arise when professionals use AI to improve efficiency without properly checking its output for hallucinations or gaps, said Maria Long, New York-based chief underwriting officer at cyber insurer Resilience. Melissa Sowa, New York-based managing director of errors and omissions product line leadership at Markel, said AI can heighten professional liability risk across many types of firms.
Legal errors are already drawing scrutiny
Law firms are seeing AI-related misconduct emerge in practice, said Laura Zaroski, Chicago-based managing director of Arthur J. Gallagher & Co.’s law firms practice. She cited cases in which attorneys filed hallucinated case law without reviewing it, sometimes resulting in monetary sanctions of $1,000 or $5,000. Those fines aren’t covered by insurance, she said.
Legal malpractice claims tied to AI remain limited, but Zaroski expects that to change. “We haven’t seen too many legal malpractice claims … I think they’re coming.”
For now, claims involving AI-assisted legal work would generally be handled like other acts, errors or omissions under legal malpractice coverage, Zaroski said. If claim frequency rises, insurers could respond with exclusions or other limits on exposure.
“They’re covered like any other act, error, or omission that you would see covered under a legal malpractice policy; it’s just like having a bad first-year associate or misreading a case,” she said. “If we keep seeing an increase in claims, we may see more carriers potentially putting in some exclusions.”
Coverage depends on policy wording and oversight
Unless a policy specifically excludes AI, professionals remain accountable for mistakes made with the tools they use, and insurers generally remain responsible for coverage, said Michael Carr, Chicago-based head of cyber portfolio underwriting, North America, for Coalition.
Many E&O and cyber policies are written broadly enough to address negligence in professional services or security and privacy failures regardless of the technology involved, said Kara Higginbotham, New York-based head of professional liability and cyber at Zurich North America. But an insurer may miss exposure during underwriting if it doesn’t understand how an organization uses or adopts AI, she said.
Markel’s standard E&O policies generally cover AI-related negligence arising from professional services, Sowa said. Its miscellaneous E&O form also lets brokers add affirmative AI coverage by endorsement, subject to a sublimit.
AI’s role in a professional error may be hard to establish after the fact. Long said that “shadow IT” makes it difficult to tell when an AI tool has been part of a service process, complicating efforts to determine how an error or omission occurred.
Choosing an appropriate tool and reviewing its work remain central controls, said Michelle Chia, New York-based chief underwriting officer for cyber in the Americas at Axa XL. Some systems are built for specific professions, such as law, and practitioners need to match the tool to the task. Using a general system that hasn’t been trained on relevant data for professional work could harm the client, she said.
Chia said an unchecked AI tool or output may be treated like relying on an untrained associate or publishing work that hasn’t been reviewed: a potential error or omission.
Competitive pressure slows policy changes
Insurers face a commercial trade-off when considering AI exclusions or added underwriting questions. Restrictive wording may reduce exposure, but it can also make a policy less attractive while competitors continue offering coverage or leave AI unaddressed.
Coalition is asking technology companies that sell AI products or services for more information than it typically seeks from technology E&O accounts, Carr said. Yet he expects insurers to avoid adding friction to other products until claims begin to emerge. Extra questionnaires can put brokers and policyholders off, he said.
Long said the soft market makes it difficult for underwriters to ask more questions, as added requirements can put an insurer at a competitive disadvantage. Sowa likewise said brokers are unlikely to accept AI exclusions if other insurers offer coverage or remain silent on AI.
If claims increase, insurers may take a firmer position through exclusions, limits or explicit coverage terms, Zaroski said. For now, AI risk is testing how well policy wording and underwriting capture firms’ actual use of these tools—and whether insurers can address that exposure without making their products harder to place.