Report from the East Bay Times
In Brief – A Federal District Judge in Illinois has partly rejected Google’s motion to dismiss a lawsuit by James Keene, a writer, TV producer and former FBI operative, alleging that Google’s AI Overviews defamed him by providing false statements to inquiries about himself. Keene claims that he found that some queries about himself made to Google Search were answered by AI responses saying he had killed three women and was serving a life term in prison. Although he had served time in prison on marijuana-related charges, the murder-related claims were false. He said that he repeatedly alerted Google about the errors, but they continued to show up for several months. Google’s motion to dismiss argued that AI Overviews are not considered statements of fact and that Keene is a “limited public figure” that needed to “plausibly plead” malice on Google’s part.
Context – AI chatbots and search are converging. And all Generative AI chatbots “hallucinate” to some degree. This issue has bedeviled some regulators, especially in Europe. Italy’s data protection authority, which banned OpenAI for a few months in 2023, argued that hallucinations violated the GDPR’s principle of data accuracy, and the European Data Protection Board Taskforce pressed OpenAI to make “explicit reference” that generated text “may be biased or made up” and has a “limited level of reliability”. Giving clear warning to users that AI responses might be in error is at the heart of the key hallucination liability case in the US, a defamation lawsuit against OpenAI in Georgia, in which the judge dismissed the lawsuit and cited the company’s “extensive warnings to users that errors of this kind could occur.” The German Regional Court of Munich recently went the other way with Google, ruling it was directly liable for erroneous statements produced by its AI Overviews, rejecting Google’s argument that users are informed that the AI answers are not always accurate and that they should verify information on their own. If GAI providers are legally held to the veracity of their answers, that will be a challenge to their viability.