Amazon's Alexa has been claiming the 2020 election was stolen

Amid concerns the rise of artificial intelligence will supercharge the spread of misinformation comes a wild fabrication from a more prosaic source: Amazon's Alexa, which declared that the 2020 presidential election was stolen.

Asked about fraud in the race - in which Joe Biden defeated President Donald Trump with 306 electoral college votes - the popular voice assistant said it was "stolen by a massive amount of election fraud," citing Rumble, a video-streaming service favored by conservatives.

The 2020 races were "notorious for many incidents of irregularities and indications pointing to electoral fraud taking place in major metro centers," according to Alexa, referencing Substack, a subscription newsletter service. Alexa contended that Trump won Pennsylvania, citing "an Alexa answers contributor."

Multiple investigations into the 2020 election have revealed no evidence of fraud, and Trump faces federal criminal charges connected to his efforts to overturn the election. Yet Alexa disseminates misinformation about the race, even as parent company Amazon promotes the tool as a reliable election news source to more than 70 million estimated users.

Amazon declined to explain why its voice assistant draws 2020 election answers from unvetted sources.

"These responses were errors that were delivered a small number of times, and quickly fixed when brought to our attention," Amazon spokeswoman Lauren Raemhild said in a statement. "We continually audit and improve the systems we have in place for detecting and blocking inaccurate content."

Raemhild said that during elections, Alexa works with "credible sources" like Reuters, Ballotpedia and RealClearPolitics to provide real-time information.

After The Washington Post reached out to Amazon for comment, Alexa's responses changed.

To questions The Post had flagged to the company, Alexa answered, "I'm sorry, I'm not able to answer that." Other questions still prompt the device to say there was election fraud in 2020.

Jacob Glick, who served as investigative counsel on the Jan. 6, 2021, committee, called Alexa's assertions nearly three years after the violent attack on the U.S. Capitol "alarming."

"If major corporations are helping to give life to the 'big lie' years after the fact, they're enabling the animating narrative of American domestic extremism to endure," said Glick, who now serves as a policy counsel at the Georgetown University Law Center's Institute for Constitutional Advocacy and Protection. "They should be doing everything they can to stop the 'big lie' in its tracks, lest we see history repeat itself."

The answers foreshadow a new information battleground in the 2024 elections, as Trump - the GOP front-runner - campaigns for the White House on the false claim that election fraud drove his 2020 loss.

Tech companies have long resisted being cast as arbiters of truth online. But technologies like voice assistants and chatbots, which serve up a single definitive answer rather than millions of ranked links or posts, stand to magnify debates about online speech that have dogged Silicon Valley since the 2016 election.

Voice assistants and advanced chatbots are only as accurate as the websites, news reports and other data they draw from across the web. These tools risk baking in and amplifying the falsehoods and biases present in their sources.

Raemhild said that Alexa draws data from "Amazon, licensed content providers and websites like Wikipedia."

Amazon founder and former CEO Jeff Bezos owns The Washington Post. The Post's interim CEO, Patty Stonesifer, sits on Amazon's board.

In recent years, Amazon's Alexa has proliferated across a number of devices. It's been embedded in inexpensive home speakers, headphones, TVs and cars - a familiar helper that sets alarms, plays songs and checks the weather for millions of Americans.

But Amazon has sought to position the voice assistant as a reliable source for information about elections for the last half decade. Ahead of the midterm elections in 2018, the company encouraged customers in a blog post to ask, "Alexa, what's my election update?"

"We believe voice provides a unique, simple, and delightful way to learn about election information, and we want to be as helpful as possible for customers when they're preparing to vote," the company said at the time.

Amazon also previously partnered with government agencies concerned about providing accurate information about civic processes. In 2020, the California Secretary of State's office created a skill that allowed voters to ask Alexa "Where is my polling place?" "What time do the polls close?" and "What are the election results?"

The company also worked with the Census Bureau to ensure that the voice assistant didn't spread falsehoods that would deter people from taking part in the once-a-decade count, which has far-reaching implications for elections and decisions about the American economy.

The voice assistant is poised to reach a wide swath of Americans before next year's election: More than 75 million people in the United States are expected to use Alexa at least once a month in 2024, according to an analysis from Insider Intelligence, a market research company.

Alexa and older voice assistants use a technological approach known as neural networks to answer a certain set of questions and do chores. Those systems function much like a phone tree or customer service line. When a customer asks Alexa a question, the speech is translated into text and then automated systems pull the most relevant information using a variety of sources.

Amazon also crowdsources answers from customers. The company says it moderates these responses with automation, trained moderators and customer feedback.

"During elections, we provide source and media outlet attribution so that customers know exactly where election information is coming from," Raemhild said.

There is limited information on how voice assistants may spread misinformation, yet some researchers argue they could be particularly effective vectors for falsehoods. Users have "higher trust" in the assistants due to their humanlike characteristics, according to a paper written by researchers at King's College London. Customers may also think the information they're getting is coming directly from the tech companies, rather than a third-party provider, making it seem more reliable, according to the paper.

Alexa - in contrast to most generative AI systems - discloses the sources of information it uses to provide answers.

The system is not always incorrect. When asked "Who won the 2020 election?" the assistant correctly answers "Democrat Joe Biden," citing election results from Reuters. Changing the phrasing of a question can elicit different responses, which are at times accurate. When asked if the 2020 election results were fraudulent, Alexa says, "There is no evidence of widespread fraud in the 2020 election - in Pennsylvania or anywhere else," referencing CNN.

The inconsistent answers from Alexa could reflect an attempt by developers to draw from a wide range of news sources across the political spectrum to address concerns of bias, said Meredith Broussard, an associate professor at New York University and author of "More Than a Glitch: Confronting Race, Gender and Ability Bias in Tech."

Developers "often think that they have to give a balanced viewpoint and they do this by alternating between pulling sources from right and left, thinking this is going to give balance," Broussard said. "The most popular sources on the left and right vary dramatically in quality."

Such attempts can be fraught. Earlier this week, the media company the Messenger announced a new partnership with AI company Seekr to "eliminate bias" in the news. Yet Seekr's website characterizes some articles from the pro-Trump news network One America News as "center" and as having "very high" reliability. Meanwhile, several articles from the Associated Press were rated "very low."

Tech companies have faced backlash in the past for similar decisions on social media. When Facebook unveiled its specialized "News" tab in 2019, media watchdogs criticized the project for including Breitbart News, a web outlet linked to right-wing causes once run by former Trump adviser Stephen K. Bannon, alongside traditional outlets, including The Post.

Since the beginning of the year, Republicans in Congress have escalated the pressure on tech companies to take a hands-off approach to misinformation, opening an investigation into long-running allegations that the industry is biased and colluding with Democrats to censor their views online.

Amazon has largely avoided the skirmishes over online speech, though its cloud services provider AWS and video-streaming service Twitch made high-profile takedowns in the wake of the Jan. 6 attacks. Rep. Jim Jordan (R-Ohio), the chairman of the House Judiciary Committee and a contender for House speaker, subpoenaed Amazon chief executive Andy Jassy in February, seeking communications about whether the executive branch "coerced and colluded" with the company to censor information online.

It's unclear how long Amazon's Alexa voice assistant has been serving up false claims in response to questions about the 2020 election. Broussard said it's also possible the answers reflect Amazon's divestment in the current version of Alexa, as the company seeks to reboot its voice assistant. (A September Washington Post review found that the new version of Alexa, a cross between the assistant and ChatGPT, repeatedly got questions wrong.)

"That's what happens to abandoned tech platforms," she said. "They get exploited by bad actors."

Alexa is an outlier in incorrectly answering whether the 2020 election was stolen. In tests by The Post, Google Home said even William P. Barr, Trump's own attorney general, says the election was not stolen, citing KCRA, an NBC affiliate. Siri serves up a list of links including KCRA, the Associated Press and a peer reviewed journal of the National Academy of Sciences disputing claims of systematic election fraud.

ChatGPT, which is powered by more advanced "large language model" technology, gives an unequivocal dispute of election theft, citing multiple audits, recounts and court rulings that affirmed the legitimacy of the election results.

There are mounting concerns from lawmakers in both parties about the role of AI in elections, especially the role of AI chatbots. Sen. Amy Klobuchar (D-Minn.) has talked about asking ChatGPT what voters should do if there were lines at a polling location in Bloomington, Minn. The chatbot responded, "Go to 1234 Elm Street." However, Klobuchar says the location does not exist.

"Think about if that happened on Election Day," she said at a recent news conference. OpenAI, ChatGPT's creator, declined to comment.

Yet despite a growing clamor in Congress to respond to the threat AI poses to elections, much of the attention has fixated on deepfakes.

However, Glick warned Alexa and AI-powered systems could "potentially double down on the damage that's been done."

"If you have AI models drawing from an internet that is filled with platforms that don't care about the preservation of democracy . . . you're going to get information that includes really dangerous undercurrents," he said.

- - -

Shira Ovide contributed to this report.