Google expands AI training to images, audio and videos users upload

Google has quietly broadened the types of user content it can use to train its artificial intelligence models, with media uploaded through several of its search-related services now forming part of the company’s growing AI dataset.

The change affects content submitted through tools such as Google Lens and Google Translate, where users upload images, audio, videos or files to complete searches or translations.

As highlighted in a report by PYMNTS, the updated policy was introduced without a major public announcement, reflecting the increasing demand among AI developers for fresh, high-quality data.

Under the revised terms, media uploaded for visual searches, voice searches or translation services may be used to improve Google’s AI systems.

However, the company has not extended this specific policy to personal content stored in Google Photos, which remains outside the scope of the latest update.

The move comes as technology companies continue competing to build more capable generative AI models, a process that relies heavily on vast amounts of training data.

As existing public datasets become more limited, firms are increasingly looking at user interactions within their own products to strengthen future AI development.

Users are automatically included in the programme, although Google provides an option to opt out. Those who do not want their uploaded media used for AI training can adjust their settings by visiting their Search Services History page and disabling the ‘Save Media’ option. Users can also review their Search Services Personalisation settings to manage what information is retained.

Another option highlighted in the report is for users who prefer not to see AI-generated summaries in Google Search. Adding “-AI” to a search query can remove AI Overview results from the search interface.

Follow Joburg ETC on Facebook, Twitter, TikTok and Instagram

For more News in Johannesburg, visit joburgetc.com

Source: PYMNTS