The AI Policies of Project 2025

Are a Perplexing Mishmash of Accusations and Inconsistencies

“There’s nothing on training, there’s nothing on privacy, there’s nothing on security and equity, none of that.”

President Joe Biden issued an executive order on artificial intelligence (AI) in October of last year. It contained a number of grand but ambiguous promises, such as developing standards for watermarking content created by AI and shielding individuals from prejudice and fraud made possible by AI.

It’s unlikely that presidential candidate Donald Trump has any profound ideas about artificial intelligence (AI) as he hardly knows how to operate a computer. However, those close to him have exposed a disorganized strategy for the IT sector that aims to dismantle Biden’s meek safeguards and replace them with a buffoonish assortment of peculiar and occasionally incoherent plans, driven by a combination of petty complaints and the financial interests of Trump’s affluent backers.

Consider the official policy platform of the Republican Party, which is written in Trump’s distinctive style of emphasis on capitalization and belligerence.

“We will repeal Joe Biden’s dangerous Executive Order that hinders AI Innovation, and imposes Radical Leftwing ideas on the development of this technology,” it yells. “In its place, Republicans support AI Development rooted in Free Speech and Human Flourishing.”

In what way may that occur? The campaign’s official policy statement, which drops all further mention of the technology after those outrageous claims, would provide no means to find out.

Further hints are contained in The Heritage Foundation’s Project 2025, a jaw-dropping document written by Trump-affiliated political operators that advocates for radical conservative changes to the political system in the event of his victory.

However, Project 2025 presents an image that is recognizable to everyone who witnessed Trump’s first term: an incoherent stew of bluster, cruelty, and xenophobia rather than providing a cogent perspective on AI.

The memo makes a number of inconsistent requests, including increasing or decreasing the use of AI in government, utilizing AI to spy on Medicare beneficiaries, enhancing espionage agencies with AI, and taking numerous broad swipes at China. The more closely you examine it, the less it seems to have any true substance or relevance to Trump’s actual views, let alone any kind of complete or even somewhat workable ideas.

Project 2025, for example, adopts the excessively generalized attitude that the US government will “stop US entities from directly or indirectly contributing to China’s malign AI goals.” However, Trump changed his mind earlier this year on the senators’ suggested prohibition on TikTok, stating that he was no longer in favor of prohibiting the video-sharing app in the US. The fact that Bytedance, a massive Chinese tech company, owns TikTok and is among the nation’s most well-known AI companies runs counter to the idea that a conservative government would or perhaps could prevent “US entities” from supporting Chinese AI initiatives.

There are more instances of hypocrisy in the document’s AI worldview. Moreover, it accuses China of using AI to advance “authoritarian control domestically and export its authoritarian governance model overseas”; on the other hand, it asserts that the government should employ AI to “better exploit” unstructured data sets and classified intelligence information, urging the US to expand the reach of its own espionage activities.

One approach to accomplish this goal is to assign the intelligence arm’s leadership to partisan loyalists, as suggested by Project 2025 among many other ideas. The main objective of the project, which has been well documented, is to consolidate authority within the executive branch and appoint loyalists while modernizing the impartial government bureaucracy.

When you combine it with AI-enhanced monitoring and espionage activities, it sounds like it’s trying to turn the US into China.

Project 2025’s stated intentions as well as its implications for artificial intelligence are frightening. The manifesto, for example, condemns what it vilely terms “abortion tourism,” which is the practice of women from areas where abortion is illegal traveling to blue states in order to receive reproductive healthcare. It does this by claiming that the “abortion surveillance” system of the Centers for Disease Control (CDC) is “woefully inadequate.

It suggests that the CDC gather far more information regarding women’s reproductive care and that the federal government limit Medicaid funds to states that don’t comply in order to address these alleged shortcomings. What’s to stop the federal government using AI to help it sort through data to find and penalize suspected abortions in this far-fetched, Christian Nationalist fantasy?

Similarly, Project 2025 has an infatuation with Medicare and Medicaid data. It suggests using artificial intelligence (AI) to “reduce waste, fraud, and abuse” in Medicare in addition to advocating for an expansion of medical data collection.

Algorithmic bias is a terrible fact of AI that may and has caused harm to minority groups. Medicare fraud is a serious concern, but so is algorithmic bias. However, the document does not even address this well-known weakness in AI technology, instead preparing to allow it to make judgments in critical areas like healthcare and medicine.

Reducing the administrative burden on AI enterprises is another major goal of Project 2025. Its persistent desire to surpass China in AI dominance and its fervor for deregulatory Silicon Valley libertarianism, which permeates Trump’s approach to both AI and crypto, appear to be contributing factors to that.

It makes the case in a single sentence that the US Patent Office ought to liberalize and streamline its processes in order to welcome AI companies. In the next, it advocates for the government to resort to commercial enterprises rather than producing costly AI models internally.

The one consistent idea in Project 2025’s disorganized AI policy is that by using tech contractors, AI should be used to increase the authority of intelligence agencies. For Peter Thiel, the billionaire libertarian who is very close to Trump’s running mate, Ohio Senator JD Vance, that is extremely convenient.

Thiel is a co-founder of Palantir, a covert data and AI company that had early investors including the FBI, NSA, and CIA. During Trump’s presidency, Thiel was dubbed America’s “shadow president”; the administration allegedly exploited Palantir’s technology to monitor and suppress political dissent, among other things.

Palantir and rival government contracting behemoth Microsoft teamed up more recently, according to Bloomberg, “in a bid to sell software, including OpenAI’s GPT-4, to US defense and intelligence agencies for top-secret tasks.” It’s important to note that Thiel is only one of several powerful Silicon Valley figures who have backed Trump in recent days.

Not just the Heritage Foundation, but other powerful Trump allies are also reshaping AI. The Washington Post reports that a new executive order being drafted by a group of Trump friends would allow the White House to implement a number of AI “Manhattan Projects” with a military and intelligence focus while easing “burdensome regulations for companies.”

The current state of the AI industry is largely unregulated, thus the concern with purported red tape is almost ridiculous. In reality, digital regulations should not agitate CEOs and impede business; rather, they should safeguard the security and privacy of US residents, establish corporate and governmental transparency requirements, and guarantee that safety stays at the forefront of technological innovation.

Nevertheless, none of the AI recommendations in these papers deal with privacy, safety, openness, or security. They also don’t talk about trying to make algorithms more equal, addressing copyright issues with AI companies, or developing criteria for AI training data.

The most striking thing is that none of the texts mention AI replacing humans in the workforce, which is ironic considering Trump and his supporters’ fixation on the notion that immigrants will steal employment.

According to Roxana Muenster, a PhD candidate in Cornell’s communication department and a Brookings COMPASS fellow, these glaring disparities show the complete misunderstanding of the scattershot strategy plan.

What particularly amazed me was how long more than 900 pages and how in-depth it is in some topics. Next, you have one of the most significant issues in tech policy, communication policy, and other areas; this will have a significant impact on many jobs, making it the most significant issue.

And there isn’t much information available about it,” Muenster said to Futurism, going into additional detail regarding Project 2025’s inadequate AI guidelines for Brookings. “There aren’t any detailed plans.”

“Of course, everyone’s kind of playing catch up [to AI policy], because the technology is faster  that’s normal,” she stated. However, “there’s nothing on training, there’s nothing on privacy, security and equity, none of that,” she went on, calling it “worrisome” that the Biden directive would be repealed in the absence of a comprehensive AI plan.

“It would just leave a huge vacuum in terms of AI policymaking,” Muenster stated, “where people are less safe as a result.”

A confusing, contradictory, and frequently broken framework for the use and advancement of AI is painted by Project 2025 and other conservative policy documents released ahead of the 2024 election. These documents center their haphazard approach to AI on a tunnel-visioned focus on the need to beat China to the AI punch without taking into account American civil liberties. The documents are rooted in fringe, dystopian isolationist-meets-technocratic ideologies. But that might be because this movement views the AI sector as a weapon to be shaped around more expansive, overarching ideological goals rather than as a necessarily regulated sector of society.

Project 2o25 is “not intended to be the best version” of AI policy, claims Muenster. “It’s intended to be an ideologically beneficial version.”

“Their tech policy really just feels like, ‘let’s tailor this in a way that we can better achieve these other goals that we have,'” she said.

Does any of this pose a real risk? It’s challenging to determine. Despite his constant stream of controversies and outbursts, Trump’s first term’s policy implementation was mediocre at best. The individuals constructing these visions of surveillance capitalism may sincerely desire them, but in order to make that happen, they’ll require Trump’s cooperation which is never assured, not even by his allies.

The best place to start could be with what Trump has been doing with the technology to get a sense of the kind of role AI would actually play in a second Trump White House.

The former president erroneously claimed earlier this month that current vice president Kamala Harris, his opponent in the 2024 race, had used artificial intelligence (AI) to create pictures of a group of people embracing her on a Michigan airport tarmac. Trump went so far as to declare that Harris should be “disqualified” and that the “creation of a fake image” should be considered “ELECTION INTERFERENCE”; however, the pictures were real and not artificial intelligence-generated.

Less than a week later, Trump himself posted a phony AI-generated picture of Harris in red, talking to a horde of figures that resembled Soviet soldiers, with a Communist flag flying overhead, on Truth Social. Later that day, he posted a fake image of Taylor Swift and declared that Swifties, Swift’s admirers, were supporting him. With an enthusiastic “I accept,” he captioned the group of images, although he didn’t say whether or not he was aware that any of the imagery was false. Reporters questioned Trump over the fake Swiftie photos, and he said he “knew nothing about them.” In addition, the former president has posted numerous AI representations of himself, including fictitious images of him praying and riding lions.

In another instance, he shared his thoughts on AI raising the possibility of nuclear war with YouTuber Logan Paul.

“Let’s see how it all works out, it’s really powerful stuff AI,” Trump said to the popular figure on the internet.

To put it another way, Trump’s personal attitude to AI has been chaotic. The former president has not only published AI photography that he finds appealing, but he has also falsely accused AI of being used in real photos that he finds objectionable. There’s merely convenience here, no meaningful throughline to follow.

However, disorder has always been a defining feature of Trump’s political career, and AI doesn’t seem to be any different. In light of the conservative movement’s drive for deregulatory measures against the AI business and the ambiguous, isolationist AI framework proposed by Project 2025, the phrase “let’s see how it all works out” seems to sum up the movement’s position on AI.