Michael Dwyer/Associated Press
Voters leave the St. Anthony Community Center polling place during the presidential primary election on January 23, 2024 in Manchester, New Hampshire.
Tech giants including Microsoft, Meta, Google, Amazon, X, OpenAI and TikTok announced an agreement on Friday aimed at reducing the risk that artificial intelligence will interfere with the 2024 election.
The tech industry’s “agreement” targets AI-generated images, videos, and audio that can mislead voters about candidates, election officials, and the voting process. However, it stops short of calling for a complete ban on such content.
And while the agreement signals the unity of a platform collectively used by billions of users, it also outlines efforts already underway, including efforts to detect and label AI-generated content. Most of them are outlined.
Concerns about how AI could be used to mislead voters and maliciously misinform those running for office are raising concerns about how AI could be used to mislead voters and maliciously misinform those running for office. It is expected to increase further in 2020. Voices that appear to have been generated by AI were used to impersonate President Biden to dissuade Democrats from voting in January’s New Hampshire primary, and to suggest that a leading candidate had rigged votes in Slovakia’s September election. It is already being used to stream audio.
“The intentional and private generation and distribution of deceptive AI election content has the potential to mislead the public in a way that jeopardizes the integrity of the electoral process,” the agreement states. “We affirm that protecting election integrity and public trust is a shared responsibility and common interest that transcends partisan interests and national borders.”
The companies announced the deal at the Munich Security Conference, an annual gathering of heads of state, intelligence, military and diplomats known as the “Davos of national defence.”
The agreement is a set of voluntary principles and commitments by technology companies. This includes developing technology to watermark, detect, and label realistic content created with AI. Evaluate the underlying models of your AI software to identify risks of abuse. Support efforts to educate the public about AI. The agreement does not specify how the promises will be fulfilled.
Work on the agreement began earlier this year and the final agreement was reached in just six weeks. Its breadth lacks the concrete, enforceable measures that many technology commentators have called for, but perhaps reflects the challenge of bringing 20 different companies on board in such a short period of time. It seems that
Microsoft President Brad Smith said the show of industry unity is an accomplishment in itself.
“We all want and need to innovate. We all want and need to compete with each other. We want and need to grow our businesses.” said. “But it is also essential that we uphold a high level of responsibility and recognize and address very real issues, including those involving democracy.”
The 20 companies signing the agreement include companies developing AI content generation tools such as OpenAI, Anthropic, and Adobe. The document was also signed by Eleven Labs, which voice cloning researchers believe is behind Biden’s fake voice. Platforms that distribute content have also signed on, including Meta, which owns Facebook and Instagram, TikTok, and Company X, formerly known as Twitter.
Nick Clegg, Meta’s global president, said the scale of the threat posed by AI requires coming together as an industry, in addition to the efforts already being undertaken by individual companies.
“All our defenses against the deceptive use of AI during elections are only as strong as our collective efforts,” he said. “Generated AI content doesn’t just stay in the silo of one platform; it moves rapidly from one platform to the next across the internet.”
The agreement focuses on transparency, education, and detecting and labeling deceptive AI content rather than removing it, and addresses the tech industry’s reluctance to police political content more aggressively. It reflects.
Right-wing critics are waging a pressure campaign in Congress and courts against social media platforms’ policies and partnerships with government agencies and academics aimed at quashing election-related falsehoods. As a result, some technology companies have backed away from these efforts. Misinformation, propaganda and hate speech in particular have all spiked in X since Elon Musk took office, researchers say.
Microsoft’s Smith said the agreement clearly distinguishes between free speech, which both companies strive to protect, and deceptive content.
“Let everyone speak for themselves. Let groups come together to speak out about the issues they care about. But let’s try to put words in someone’s mouth that were never spoken. , do not deceive or fundamentally deceive the public,” he said. “In our opinion, that is not freedom of expression. It is what we call fraud and deception.”
New wrinkles appear on old problems
How disruptive AI will be in this election cycle remains an open and unanswered question.
Some experts worry it’s hard to overstate the risks.
“The power provided by new technologies available to adversaries will be frightening,” said Joe Kiniley, chief scientist at Free and Fair, an open-source election technology company. “I don’t think we can do science fiction writing right now that approaches some of the things we’re going to see next year.”
But election officials and the federal government insist the impact will be further limited.
The Cybersecurity and Infrastructure Security Agency, the Department of Homeland Security’s division responsible for election security, said in a recent report that generative AI capabilities are “unlikely to pose new risks, but may pose existing risks to election infrastructure.” “This could amplify the risks.” Voting Process and Cybersecurity Threats.
AI dominated much of the conversation at the Secretary of State’s meeting in Washington, D.C., earlier this month, as election officials were quick to point out that they have been battling misinformation about the election process for years. Recent advances in AI are just an evolution. It’s something they’re already familiar with.
Democratic Arizona Secretary of State Adrian Fontes said, “AI is not the great, mysterious, world-changing monster that some people are touting, but its amplifier. need to be exposed.” “It’s a tool that can spread bad messages, but it’s also a tool that allows you to discover great efficiencies.”
Sam Volante/Cronkite News
Arizona Secretary of State Adrian Fontes speaks to high school students about voter registration at the Phoenix Coding Academy on September 19, 2023.
One specific concern that frequently came up was that skepticism could encourage the public to be skeptical of what they see online, without leading to widespread distrust or disengagement from all information. It was about how difficult it was.
For example, officials expect candidates to increasingly claim that true information is generated by AI, a phenomenon known as the “liar’s dividend.”
“It’s going to be easy to claim that everything is fake,” Adriana Stephan, an election security analyst at CISA, said during a panel discussion on AI at the conference.
Regulators also pay attention to guardrails
Many of the signatories to the new technology agreement have already announced initiatives that fall within the areas covered by the agreement. Meta, TikTok, and Google require users to disclose when they post realistic AI-generated content. TikTok has banned AI fakes of celebrities used for political or commercial endorsements. OpenAI does not allow its tools to be used for political campaigning, creating chatbots that impersonate candidates, or discouraging voting.
Meta announced last week that it will begin labeling images created with leading AI tools in the coming months using invisible markers being developed by the industry. Meta also requires advertisers to disclose their use of AI in ads about elections, politics, and social issues, and prohibits political advertisers from using its proprietary generative AI tools to create ads. are doing.
Efforts to identify and label AI-generated audio and video are still beginning, even though AI-generated audio and video have already been used to mislead voters, such as in New Hampshire. I just got it.
But tech companies are developing ever more advanced technology, even as they respond to pressure about the potential for their products to be misused. OpenAI on Thursday announced a tool that generates realistic videos of up to a minute in length from simple text prompts.
The move by companies to voluntarily curb their use of AI comes as regulators struggle with how to set guardrails around new technology.
European lawmakers are poised to approve the Artificial Intelligence Act in April, a far-reaching set of rules billed as the world’s first comprehensive AI law.
In the United States, a series of federal bills regulating the technology have received little attention, including banning deceptive deepfakes in elections and creating a new agency to oversee AI. State action is accelerating, with lawmakers in about 32 states introducing bills to regulate deepfakes in elections since the beginning of this year, according to the progressive advocacy group Public Citizen.
Silicon Valley critics argue that while AI amplifies existing threats to elections, the risks posed by the technology are broader than companies’ latest tools.
“However, the main technology-related threat to this year’s elections is not the creation of content using AI, but a more familiar cause: the distribution of false, hateful, and violent content through social media platforms. researchers at the new institute said. York University’s Stern Center for Business and Human Rights wrote a report this week criticizing content moderation changes made at Meta, Google, and X.