Google, OpenAI Say US Law Lets Them Train AI on Australian Content Without Seeking Permission

Google and OpenAI have informed a federal parliamentary inquiry that their AI models are trained in the United States under American ‘fair use’ rules, bypassing the need for permission from Australian copyright holders. Roxanne Carter of Google stated that using freely available, lawfully accessed, and non-opted-out online content is how they train, as prior authorization for everything is deemed impractical. This approach leverages US copyright law for training data.

• Google and OpenAI train AI models in the US.
• They rely on American ‘fair use’ rules for training data.
• Australian content is used without seeking permission from copyright holders.
• Google’s Roxanne Carter cited the impracticality of obtaining prior authorization, estimating it would take 200 years.
• The companies use content that is freely available online, lawfully accessed, and not opted out of.