AI hasn't created new risks in film production, just new ways into old ones, BFL's Lindner says

BFL Canada's Nellie Lindner says underwriters won't automatically cover AI use, and a human checking the output is the non-negotiable

AI hasn't created new risks in film production, just new ways into old ones, BFL's Lindner says

Professional Risks

By Branislav Urosevic

Artificial intelligence has not introduced a new category of risk into film and television production. It has opened a new route to the risks the industry has always insured against – copyright, libel, title and music infringement – and underwriters are responding by asking far more questions before they will cover it, according to Nellie Lindner, national practice leader for films and entertainment at BFL Canada.

Lindner said the exposure at the heart of AI use is the same one that has always sat under errors and omissions cover.

"It was a human versus, or a writer versus, now it's a computer or AI technology," she said. What has changed is the uncertainty around ownership, copyright and liability when a producer uses a generative tool to create a script, an image or a piece of music – and whether a third party already owns what the tool produced.

Because of that, AI is never simply included in a policy, she said. Canada has only a small pool of underwriters who write production errors and omissions coverage, and they approach it consistently: coverage is not automatic, and every use has to be interrogated.

"They will not automatically provide AI coverage," Lindner said. When an E&O application comes in and a client asks whether AI is covered, the questions start. "How are you using it? What are the controls in place?"

The non-negotiable, she said, is a person standing between the tool and the finished work. Underwriters want assurance that someone is reviewing AI output to confirm it does not infringe on material owned by someone else.

"They always want to make sure that there's a human element as well, so that there is the checks and balances," Lindner said – someone confirming no third party owns the IP behind an image or a story.

She shared a case that showed how fast a submission can fall apart without that diligence. A client sent in sample images they wanted to use, and both the underwriter and Lindner checked them. The answer was immediate and definitive. "Absolutely not." The reason was simple: the images were traceable. "Copy and paste them into the reverse image search and they came up," Lindner said.

The scrutiny is not confined to one medium; it extends across all elements of a production, Lindner said, from music and scripts to titles. “It covers everything,” she said. A song generated through an AI prompt can surface lyrics that are already owned. The same logic applies to titles, with underwriters requiring clear evidence of ownership and proper clearance at every stage.

What underwriters are looking for, she said, is not additional premium but evidence of control. A client who can show how the tool is used, on what platform, under what employee guidelines, and how that use is managed and monitored, will generally get coverage. Guidelines alone are not enough. "Guidelines are one thing, but you also have to make sure that there is governance in place," Lindner said, and that it keeps pace as the technology evolves.

That, she said, is the notable change in how underwriters are approaching AI: rather than pricing around the uncertainty, they are underwriting it directly. When they do include AI, "they are doing a lot more underwriting as opposed to wanting to add on additional premiums," Lindner said – and if the controls meet their expectations, coverage is not automatically declined.

The deeper reason for the caution, she said, is that the exposure is still unproven. No one knows how large the AI claims will be until a significant one lands, and so far it hasn't. The risk, in her view, is that some clients are not disclosing their AI use at all, which leaves insurers exposed to something they never priced.

For now, Lindner said, AI sits where other emerging perils once sat – a moving target that underwriters may one day fold in automatically, but not yet. The exposure itself, stripped down, is old and familiar, and so is the fix.

“At its core, the exposure is the use of content sourced from elsewhere that may belong to someone else,” she said. "It comes down to making sure that you're verifying that ownership. And if it does belong to someone else, that doesn't mean you can't use it. You just have to get permission to do so."

Related Stories

Keep up with the latest news and events

Join our mailing list, it’s free!