Google training Bard with scraped web data? Here’s everything you may want to know

economictimes.indiatimes.com:

Google has confirmed that it is utilizing scraped web data to train its artificial intelligence (AI) systems, according to an update to its privacy policy. The search giant stated that its AI services, including Bard and Cloud AI, may be trained using publicly available data obtained from the web. Google spokesperson Christa Muldoon emphasized that the company incorporates privacy principles and safeguards into the development of its AI technologies, in line with its AI Principles.

The revised privacy policy, effective from July 1, 2023, now explicitly states that Google uses information to enhance its services and develop new products and technologies for the benefit of users and the public. It specifies that publicly available information can be utilized to train Google's AI models and build various products and features, such as Google Translate, Bard, and Cloud AI capabilities. However, the policy does not detail how Google will prevent copyrighted materials from being included in the data pool used for training.

Many publicly accessible websites have implemented policies banning data collection or web scraping for training large language models and other AI tools. The implications of Google's approach will be significant, considering global regulations like the General Data Protection Regulation (GDPR), which protect individuals from the misuse of their data without explicit consent. The origin of training data used by generative AI systems, like OpenAI's GPT-4, has become a contentious issue.