Overview
When Google unveiled its Bard AI chatbot in February 2023, the company posted a promotional GIF showing Bard answering a question about new discoveries from the James Webb Space Telescope. Bard replied that the telescope "took the very first pictures of a planet outside the Earth’s solar system." Reuters and NASA later noted that the first exoplanet images were actually taken in 2004 by the European Southern Observatory’s Very Large Telescope. The error, spotted before a live presentation, contributed to Alphabet’s shares falling 9%, wiping out about $100 billion in market value. Analysts said the misstep underscored the risks of deploying generative AI without rigorous fact‑checking.
What Went Wrong
Bard was not sufficiently fact‑checked before being used in marketing. The model generated an incorrect but plausible‑sounding fact, and the company publicised it without verifying accuracy.
How It Was Fixed
Google said it would launch a Trusted Tester program and combine external feedback with internal testing to ensure Bard’s responses are accurate. The company paused broad deployment and later rebranded its chatbot under the Gemini name with improved safeguards.