Blog

Google has announced that it is modifying its privacy policy and framework for training its AI models. This means that it will now be able to crawl through publicly available data and use it to train its models. 

This is in line with the trend of top brands—including Google itself—releasing AI-focused products and information that have the potential to change the way we work. In order to do so, Google and other companies like OpenAI need large datasets to train their chatbots or large language models in general.

As spotted by Engadget, Google has reworked the wording of its privacy policy, and switched “AI models” for “language models,” and now, it can use publicly available information to build feature sets and even complete products like Google Bard, and more. Simply put, anything and everything available online can now be used to train its AI models like PaLM 2 and in future, even Gemini.

Google said that it “uses information to improve our services and to develop new products, features, and technologies that benefit our users and the public.” And, citing examples, Google claims that “we use publicly available information to help train Google’s AI models and build products and features like Google Translate, Bard, and Cloud AI capabilities.”

Search Engine Journal notes that AI chatbots—including Google’s Bard and OpenAI’s ChatGPT could be evaluating and “reusing people’s posts, reviews, and other online content.” The publication also claims that in the future, as more generative AI products become publicly available, there will growing lawsuits stemming from training AI models without permission, and using original content.