In a Law360 article titled, “OpenAI’s Sora Backlash Shows IP Challenges For Tech Cos.,” Kirk Sigmon said OpenAI and other AI developers are in a challenging position because their models require vast amounts of material to function once training data goes into an AI model, it’s hard to retroactively remove it.
“The likelihood of them being able to retroactively go, ‘OK, we’ll pull out the training data that involves, say Mickey Mouse,’ is basically limited,” Kirk said. He noted that it is possible to add filters to prevent certain outputs, which is a common practice with AI platforms.
You can read the full article by clicking here.