Two days after President Joe Biden signed a sweeping executive order on artificial intelligence last year, Vice President Kamala Harris presented the strange document to a global AI summit. She told an international audience what sets the U.S. apart in its approach to AI safety.
At an event meant to highlight the potential catastrophes that futuristic forms of AI will bring, Harris made waves by zeroing in on current concerns: the need to quickly lock in protections without stifling innovation.
“If a senior citizen is kicked off their health insurance because of a flawed AI algorithm, isn’t that existential for them?” Harris told a crowd in London last November. “If a woman is threatened by an abusive partner with explicit deepfake photos, isn’t that existential for her?”
Now she’s running for president, and her primary opponent, former President Donald Trump, has said he wants to “cancel” the Biden order. Trump’s running mate, Sen. J.D. Vance of Ohio, also brings his own views on AI, influenced by his ties to some Silicon Valley figures who are pushing to limit AI regulation.
The growing visibility of AI in everyday life has made it a hot topic of conversation, but it hasn’t yet elevated it to a top priority for American voters. But this could be the first presidential election in which candidates develop competing visions for how to guide American leadership in the rapidly evolving technology.
These are the AI performances of the candidates:
Trump’s approach
Biden signed his AI executive order on Oct. 30 of last year, and Trump soon after indicated on the campaign trail that he would stop it if re-elected. His pledge was memorialized in the platform of this month’s Republican National Convention.
“We will repeal Joe Biden’s dangerous executive order that stifles AI innovation and imposes radical leftist ideas on the development of this technology,” Trump’s platform reads. “Instead, Republicans support AI development that is rooted in free speech and human flourishing.”
Trump’s campaign did not respond to requests for further details.
Trump didn’t spend much time talking about AI during his four years in office, though he became the first to sign an executive order on AI in 2019, directing federal agencies to prioritize research and development in the area.
To that end, tech experts urged the Trump-era White House to adopt a stronger AI strategy to match what other countries were pursuing. In 2017, not long before Google quietly unveiled a research breakthrough that helped lay the groundwork for the technology now known as generative AI, then-Treasury Secretary Steven Mnuchin dismissed concerns about AI replacing jobs, saying the prospect was so far in the future that “it’s not even on my radar screen.”
That perspective later shifted, when Trump’s top technology adviser told business leaders in 2018 that AI-driven job displacement was “inevitable” and that “we can’t sit back and hope that the market will eventually solve it.” The 2019 order called on federal agencies to “protect civil liberties, privacy, and American values” when deploying AI technologies and to help workers acquire relevant skills.
Trump also signed an executive order in the final weeks of his administration to promote the use of “trustworthy” AI in the federal government. Those policies have continued into the Biden administration.
Harris’s approach
The introduction of ChatGPT nearly halfway through Biden’s presidential term made it impossible for politicians to ignore AI. Within months, Harris summoned the heads of Google, Microsoft and other tech companies to the White House, a first step on a path that led leading developers to make voluntary commitments to ensure their technology doesn’t jeopardize people’s rights and safety.
Then came Biden’s AI executive order, which used Korean War-era national security powers to investigate high-risk commercial AI systems but was primarily aimed at protecting government use of the technology and setting standards that could spur commercial adoption. Unlike the European Union, however, the U.S. still lacks broad rules on AI — something that would require Congressional approval.
Harris already brought a deep understanding of Silicon Valley to the White House, having grown up and worked in the San Francisco Bay Area and later served as California’s attorney general, where she built relationships with some of the tech industry’s leading figures, said Alondra Nelson, former director of the White House Office of Science and Technology Policy.
Even before ChatGPT, Nelson led the White House’s efforts to craft a blueprint for an AI “bill of rights” to protect against the technology’s potential harms. But it was Harris’s speech at the Global Summit on AI Safety in London where she pulled all those threads together and “made it clear to the world what the U.S. AI strategy was,” Nelson said.
Harris said she and Biden “reject the false choice that suggests we can either protect the public or advance innovation.” And while Harris acknowledged the need to consider existential threats to humanity, she emphasized “the full spectrum of AI risks.”
“She really opened the door to the conversation about the potential risks and dangers of AI,” Nelson said.
Vance and the VCs
Trump’s choice of former venture capitalist Vance as his running mate added a new element to the campaign’s differences. So did Trump’s new endorsements from a group of AI-focused tech leaders, including Elon Musk and venture capitalists Marc Andreessen and Ben Horowitz.
Vance has acknowledged that there are some harmful AI applications, but said at a Senate hearing in July that he worries that those concerns “justify some preemptive over-regulation efforts that would just plainly entrench the established tech companies that we already have.”
Andreessen, who serves on the board of directors of Meta Platforms, has criticized a provision in Biden’s order that requires the government to audit the most powerful and ostensibly risky AI systems if they can perform a certain number of mathematical calculations per second.
In a podcast highlighting his business partner Horowitz, Andreessen said he was concerned about “the idea that we’re going to deliberately handicap ourselves with onerous regulations, and the rest of the world and China are angry about this too.”
Horowitz read aloud the RNC’s call to rescind Biden’s order, saying, “That sounds like a good plan to me.” He also noted that he and Andreessen had discussed the proposals with Trump over dinner.
Trump met with another group of VCs in a video podcast in June and shared their view that AI leadership will require vast amounts of electricity — a perspective he shared again on the RNC stage, where he said it will require “twice as much electricity as is available in our country today.” It was his only mention of AI in the 92-minute speech.
Are they that different in terms of AI?
Much remains unknown, including the extent to which Harris or the Trump-Vance ticket will heed the opinions of their competing wings in Silicon Valley.
While rhetorical differences grow, “there are a lot of similarities” between the way the Trump and Biden administrations approached AI policy, said Aaron Cooper, senior vice president of global policy for BSA The Software Alliance, which advocates for software companies like Microsoft.
Voters haven’t heard many details about how a Harris administration or a second Trump administration would change that.
“What we’re going to continue to see as technology evolves and new problems arise, regardless of who’s in the White House, they’re going to be looking at how do we get the most good out of AI while minimizing the most harm,” Cooper said. “That sounds obvious, but it’s not an easy calculation.”