- Posted on: Mar 29 2019
AI’s ‘Uncharted Territory’
Rolling out an artificial intelligence system is a bit more difficult than booting up software and saying, “Go.” For one, AI requires training—data that provides the basis for the AI system to further perfect itself. This training can run into all sorts of issues if not managed correctly, such as the increasingly popular “garbage in, garbage out” adage and bias creeping into the system. But as some photographers are learning, training AI systems can result in intellectual property questions as well.
Last week, NBC News reported that IBM has been taking images from photo-sharing website Flickr to train its facial recognition technology. What they didn’t do, though, is actually tell the photographers their images were being used. Naturally, the photographers were none too happy about the situation, but privacy experts who spoke with Legaltech News agreed on one thing: They may not have much of a leg to stand on for a copyright claim.
“Even if they did have a copyright claim, IBM would have a decent fair use claim, another issue they may rely on,” said Dunlap Bennett & Ludwig partner David Ludwig. “IBM may come out and say, ‘We are not copying the photos to sell them, we are using this to create an amazing AI tool.’”
Attorney Carolyn Wright added that most photographers don’t register their photos with the U.S. Copyright Office, which allows them to only recoup actual damages—a license fee usually under $100 and provable profits earned by the company from their image. So the question then becomes, especially with privacy front and center in the current public conversation: What risk are companies willing to assume for this data?
There’s no easy answer, said attorney Sara Hawkins, as these problems are “uncharted territory.” She added, “It’s really important. We have that data, but the question is how do we allow that data to be accessed and do we allow that data to be accessed for free?”
—Zach Warren, Law.com
Posted in: Announcements