The Law Is Coming for AI—But Maybe Not the Law You Think
Courts historically have shown sympathy to technology companies on copyright issues. Privacy, however, is another matter.
While the approval of the AI Act in the European Parliament on Wednesday will no doubt go down in history as a day of reckoning for generative artificial intelligence, it was not the first. That honor belongs to March 31, when, citing a lack of compliance with various European data protection provisions, Italy’s data protection authority banned OpenAI’s ChatGPT.
Products like ChatGPT and Google’s Bard are members of the generative pre-trained transformer family of language models. These models rely on huge training data sets, to which they apply various artificial intelligence and machine-learning techniques, supervised and reinforced by human feedback, including user interactions with the model. Much of the training data is obtained by mining publicly available content, accessing data repositories and scanning large open data sets.
Just because these data sets are publicly available, however, does not mean the data itself is legally available to use.