Google Bard will respond to questions from users using information found from the internet, and will also have a function to allow users to search for suggestions for queries on Google’s eponymous search engine.

Sources for information will also be displayed by Google Bard, with Google warning that due to the AI model used, there was a risk that the tool could reflect real-world biases and stereotypes.

The tool is powered by a research large language model (LLM) called LaMDA, which will be updated with newer models over time.

Google said that there was a risk the model could provide inaccurate, misleading or false information while presenting it confidently, but would also provide multiple drafts of its responses to allow users to select the best starting point.

There are plans to expand the trial to more countries and languages as it develops.

In a blog post about Google Bard’s release by vice-president, product Sissie Hsiao and vice-president, research Eli Collins, the company said it would continue focusing on developing the tool’s quality and safety during this experimental phase.

“We’ll continue to improve Bard and add capabilities, including coding, more languages and multimodal experiences,” Collins and Hsiao wrote.

“And one thing is certain: we’ll learn alongside you as we go. With your feedback, Bard will keep getting better and better.”