OpenAI has launched a new product called SearchGPT, which it hopes will rival Google.

The product uses AI models similar to the company’s more famous ChatGPT. However, it aims to provide “fast and instant answers from clear and relevant sources” by including real-time information.

The company emphasized that the product is still in its early testing phase and appears to have made an error in its announcement.

A user asked about music festivals in a city in North Carolina in August during a demo video. The Atlantic reported that the system responded but incorrectly stated the dates, telling users the concerts would be in August when the box office would actually be closed.

OpenAI told The Atlantic, “This is an early prototype, and we will continue to improve it.”

This error follows noticeable mistakes made by similar AI systems. Last year, when Google launched its tool, known then as Bard and later rebranded as Gemini, there was an error in the promotional video claiming that an image of an exoplanet was taken for the first time by a NASA telescope.

Both errors seemed to stem from the systems’ inability to fully understand the data they relied on. SearchGPT’s dates were real but taken out of context, while Google’s Gemini appeared to confuse the first-ever exoplanet image with the first taken by a NASA telescope.

AI researchers have warned that such errors are common because AI systems struggle to understand context. These systems are also prone to “hallucinations,” where they provide factually incorrect answers with authoritative-sounding language.