Earlier this week, we covered the ethical and infringement concerns raised by the tech community over big companies using transcribed YouTube videos to train their AI models without obtaining the consent of the content creators. OpenAI, Meta, and Google received criticism for violating the rules regarding the independent use of YouTube videos. Apple also recently found itself in hot waters for the claim that it steals content for its OpenELM model.  The Cupertino tech giant has now responded to the controversy by presenting its side of the story and clearing the air regarding unethical practices for training its LLM model.

Apple denies speculations regarding unethical AI practices by stating that Apple intelligence was trained on licensed content and not stolen YouTube videos

Google, Meta, and OpenAI have received backlash for using subtitles harvested from more than 170,000 videos of well-known YouTubers to train their AI models. An earlier report highlighted that Apple also uses transcribed YouTube content for its OpenELM model, joining other companies in the highlighted ongoing unethical AI practices. The company has now come to its defense and shed some light on the matter.

As reported by 9toMac, Apple has confirmed to the channel that the OpenELM model is not linked to its other AI initiatives. Apple Intelligence and its LLM models are trained through licensed data. Apple has explained to users and the tech community that the OpenELM was part of a research initiative, and the company used the Pile dataset to train its open-source model. It was created to showcase its open language model development to the public by making it readily available on Apple’s Machine Learning Research site.

Apple further clarifies that the OpenELM model released in April is not linked to Apple Intelligence or its AI-powered features. The tech giant also expressed that it had no plans to release any versions of the OpenELM and that it was specifically a research contribution. However, Apple Intelligence is claimed to rely on completely ethical practices for training, with millions being paid to publishers and for licensed data. Apple explained this in detail in a research paper and the heed it places on responsible AI development.

Big companies, however, have been using the YouTube Subtitles datasets of the non-profit organization EleutherAI to train their AI models and raising serious concerns regarding permission, ethical AI, and copyright infringements. Even though Apple Intelligence is not part of the ongoing controversy, all the big companies should be more transparent with their material extraction techniques and AI training methods to avoid becoming part of such issues.