Google announced a major expansion of its AI chatbot tool, giving it the ability to access users' personal information from Google Workspace apps like Gmail, Docs and Drive. While promising greater convenience and requiring users to opt-in, the integration also raises concerns about privacy.
Google says it has taken privacy and security into consideration before rolling the feature out.
New Bard Extensions allow the chatbot to search within a user's Google accounts to provide personalized responses tailored to their data.
"With Extensions, Bard can find and show you relevant information from the Google tools you use every day — like Gmail, Docs, Drive, Google Maps, YouTube, and Google Flights and hotels," said Yury Pinsky, Director of Product Management for Bard, in an official blog post.
For instance, Bard could draft a cover letter using details from a resume stored on Google Drive. Or it could scan your inbox to find travel details and automatically gather flight/hotel info into a trip planning document. Instead of using technical words and instructions, users can ask Bard in natural language to look for specific information like, “What did Anna say about our product meeting last Saturday?” Or, “What did John Doe find in his experiment?” Google will reply based on the contents of your emails or documents stored in Google Drive.
"It's the first time a language model product is truly integrating with your personal data," Jack Krawczyk, Bard's product lead, told The Verge.
Google confirms that integrating AI with your data does not mean Google is using your data to train its large-language model (LLM)—it’s just using it as input to provide a result generated using an already trained AI model.
Google states that the choice of using its extensions is optional, and user content "is not seen by human reviewers [or] used by Bard to show you ads."
Beyond data access, Google is also expanding Bard's ability to double-check its responses using the "Google it" button. This will highlight whether Bard's answers are validated or contradicted by Google Search results, which could help mitigate the effects of model hallucination.
Among other updates, tools like Google Lens image uploads are now supported in 40+ languages too.
This move comes just after the tech giant is adapting to the new AI era. Google itself has also updated its SEO guidelines, as reported by Decrypt, to acknowledge AI-generated content rather than solely focus on "content written by people." This shift recognizes the growing role of AI in content creation, even though low-quality AI content may still be penalized.
As machines continue to replicate human skills, we may want to occasionally "Google" whether we can still tell the difference.