Schiff Demands Answers from Google Regarding Misleading AI Overview Feature
Today, Representative Adam Schiff (D-Calif.) demanded Google CEO Sundar Pichai address concerns over the company's new AI Overviews feature, which has been providing users with misleading political and medical information. Schiff, in the letter, also seeks clarity on the steps Google is taking to ensure the feature prioritizes reputable sources of information.
“The AI Overviews feature is intended to provide an AI-generated snapshot with key information accompanied by web links. However, reporting has already found the feature to provide misleading information from a wide range of untrustworthy sources. These sources often include satirical sites, or at worst, long-debunked conspiracy postings,” wrote the lawmakers.
“We appreciate Google’s efforts to fix these errors by improving the feature through continuing tests, taking down responses that violate the company’s content policies, and labeling AI-generated responses as experimental. However, more must be done to adequately ensure that Google can remain a trustworthy and reliable source of information for Americans,” continued the lawmakers.
The lawmakers also included a list of eleven questions to ensure Google prioritizes reputable sources of information for which they demand answers by August 1, 2024.
The letter was also signed by Representatives Henry C. "Hank" Johnson, Jr. (D-Ga.), Donald S. Beyer Jr. (D-Va.), Lori Trahan (D-Mass.), and Pramila Jayapal (D-Wash.) and is supported by Public Citizen.
Read the full letter and list of questions here.
Schiff has been a leader on the regulation of Artificial Intelligence. Last year, Schiff backed the Federal Election Commission's (FEC) unanimous decision to advance Public Citizen’s petition, pushing for rulemaking to address the use of generative artificial intelligence (AI) in deceptive campaign advertisements. In April, Schiff introduced the Generative AI Copyright Disclosure Act, demanding transparency from companies regarding their use of copyrighted work to train their generative AI models.
###