Meta Platforms Inc (NASDAQ:FB) utilized public user Facebook and Instagram posts to train the new Meta AI virtual assistant unveiled at its Connect developer conference this week, according to reporting from Reuters.
Meta's president of Global Affairs, Nick Clegg, insisted the company did not train its AI on content set to private and said the company worked to omit private details from the datasets it did use.
For example, Clegg said that Meta AI was not trained on any LinkedIn content.
"We've tried to exclude datasets that have a heavy preponderance of personal information," Clegg said.
Meta also included safety restrictions on what the AI could generate, including a ban on the creation of photorealistic images of public figures.
Meta AI is a conversational assistant coming to Messenger, WhatsApp and Instagram in the coming months. It will be able to deliver real-time information and generate photo-realistic images based on user-inputted prompts.
The AI will be able to respond using more than two dozen AI chatbot personalities, including Meta celebrities such as Bear Grylls, Chloe Kim and Josh Richards, among others.
Meta AI was made using the Llama 2 large language model that Meta released for commercial use in July.
Despite the precautions taken by Meta, Clegg said he still expects a “fair amount of litigation” related to whether the use of copyrighted materials in creating AI content is considered fair use.
"We think it is, but I strongly suspect that's going to play out in litigation," Clegg said.
Shares of Meta rose more than 2% Thursday.
Contact Andrew Kessel at andrew.kessel@proactiveinvestors.com
Follow him on Twitter @andrew_kessel