X users blasted the chatbot for the mistakes, with one calling AI a disaster for real time events. — AP
When a massive 8.8 magnitude earthquake struck off Russia's Pacific coast on Tuesday, one core worry immediately emerged: A life-threatening tsunami. Around the Pacific Ocean, weather authorities leapt into action, modelling the threat and releasing warnings and advisories to prepare their communities for what could be a horrific night. As the news unfolded, residents of Hawaii, Japan and North America's West Coast jumped onto their devices to seek evacuation plans and safety guidance.
Online, some got their information from artificial intelligence chatbots. And in the moment of potential crisis, a few of those newly prevalent tools appear to have badly bungled the critical task at hand.
Grok, the chatbot made by Elon Musk's Bay Area-based xAI and embedded in the social media site X, repeatedly told the site's users that Hawaii's tsunami warning had been cancelled when it actually hadn't, incorrectly citing sources. Social media users reported similar problems with Google Search's AI overviews after receiving inaccurate information about authorities' safety warnings in Hawaii and elsewhere. Thankfully, the tsunami danger quickly subsided on Tuesday night and Wednesday morning without major damage.
Still, the issues speak to the growing role of AI tools in people's information diets – above other search results, Google often places AI overviews that compile information from around the Internet, and Grok, the chatbot, automatically responds to users that tag "@Grok" on X posts – and to the tools' potentially dangerous fallibility.
On Tuesday evening, Grok wrote that the National Weather Services's Pacific Tsunami Warning Center had cancelled Hawaii's tsunami warning at 3.23pm local time "after data showed no significant waves. No impacts occurred – Hawaii is safe." This wasn't true. Authorities had issued the tsunami warning at 2.43pm, and waves didn't reach the islands until later in the evening.
In reality, the Honolulu-based weather authority kept its warnings active until after 10.30pm, when it downgraded the guidance to an advisory. (Grok wasn't wrong everywhere; just four minutes after its "3.23pm" claim, it accurately told another user that "evacuations are ongoing" and that warnings were active for Hawaii and parts of Alaska.)
X users blasted the chatbot for the mistakes, with one calling AI a "disaster for real time events." Another user tagged Musk and included several screenshots of Grok's errors, writing that the chatbot had "spread life-endangering misinformation during the tsunami event."
A critic of Google – who prompted the search tool to show an AI overview by adding "+ai" to their search – called the text that showed up "dangerously wrong."
X did not respond to SFGATE's request for comment. Google, which dubs high-stakes queries "Your Money or Your Life," said in a May document explaining the AI overviews that such queries give the tool "an even higher bar for showing supporting information from reliable and trustworthy sources."
Google spokesperson Ned Adriance wrote, "Search quickly and automatically updates based on fresh web content and the latest information. In this case, Search now shows a dedicated alert feature that provides up-to-date information for people searching about this topic." ("Tsunami" searches on Wednesday morning by this reporter yielded a small "advisory" symbol and then, once the advisory was cancelled in the Bay Area, just a selection of news stories.)
Grok, in reply to one of the posters complaining about its errors, wrote, "We'll improve accuracy." – SFGate, San Francisco/Tribune News Service
