The National Institute of Standards and Technology (NIST) has changed the language guiding scientists who work with the US Artificial Intelligence Safety Institute (AISI). The updated agreement drops several familiar AI governance priorities and adds a new emphasis on ideological bias, human flourishing and economic competitiveness.
The shift matters because AISI was created to study risks from powerful AI models. The new wording does not end the institute, but it changes the stated skills and priorities expected of researchers in its consortium.
What changed in the AISI agreement
The update appears in a cooperative research and development agreement for AI Safety Institute consortium members, sent in early March. According to the source article, NIST’s new instructions remove references to “AI safety,” “responsible AI,” and “AI fairness” from the skills it expects of members.
The agreement also adds a request that researchers prioritize “reducing ideological bias, to enable human flourishing and economic competitiveness.” That language closely tracks a broader Trump administration focus on AI systems that are not shaped by what it describes as ideological bias or engineered social agendas.
Earlier language encouraged technical work that could help detect and address discriminatory model behavior connected to gender, race, age or wealth inequality. The source article notes that those forms of bias can affect users directly and can disproportionately harm minorities and economically disadvantaged groups.
The new agreement also removes references to building tools “for authenticating content and tracking its provenance” and “labeling synthetic content.” That change points to a weaker stated focus on tracking misinformation and deep fakes.
Competitiveness moves to the center
The update does not only remove language. It also adds a more explicit America-first framing. One working group is asked to develop testing tools “to expand America’s global AI position.”
That emphasis fits with public statements from the Trump administration. An executive order issued by the Trump administration this January revoked the Biden administration’s order but kept the AI Safety Institute in place. The order states: “To maintain this leadership, we must develop AI systems that are free from ideological bias or engineered social agendas.”
Vice president JD Vance made a similar point at the AI Action Summit in Paris in February, saying the US government would prioritize American competitiveness in the race to develop and benefit from AI. “The AI future is not going to be won by hand-wringing about safety,” Vance told attendees from around the world. The US delegation to the event did not include anyone from the AI Safety Institute.
That creates a sharp contrast with the institute’s original mission under Biden. The AI Safety Institute was created by an executive order issued by the Biden administration in October 2023, during a period of heightened concern about rapid AI progress. Under Biden, the institute was tasked with examining risks such as cyberattacks, chemical or biological weapons, and whether advanced models could become deceptive and dangerous.
Researchers warn about what may be lost
Some researchers connected to AISI see the new language as more than a branding change. One researcher at an organization working with the AI Safety Institute, who asked not to be named for fear of reprisal, said: “The Trump administration has removed safety, fairness, misinformation, and responsibility as things it values for AI, which I think speaks for itself.”
That researcher warned that ignoring these issues could hurt regular users if algorithms that discriminate based on income or other demographics are allowed to go unchecked. The researcher added: “Unless you're a tech billionaire, this is going to lead to a worse future for you and the people you care about. Expect AI to be unfair, discriminatory, unsafe, and deployed irresponsibly.”
Another researcher who has worked with the AI Safety Institute in the past questioned the new framing. “It’s wild,” the researcher said. “What does it even mean for humans to flourish?”
Stella Biderman, executive director of Eleuther, a nonprofit working with the AI Safety Institute, connected the rewrite to the administration’s broader direction. “Those changes are pretty much coming straight from the White House,” she said. “The administration has made its priorities clear, [and] it isn't surprising to me that rewriting the plan was necessary to continue to exist.”
The political backdrop around AI bias
The article places the new agreement inside a wider political fight over AI bias. Elon Musk, who is currently leading a controversial effort to slash government spending and bureaucracy on behalf of President Trump, has criticized AI models from OpenAI and Google. Last February, he posted a meme on X in which Gemini and OpenAI were labeled “racist” and “woke.”
Musk often cites an incident involving a Google model debating whether misgendering someone would be wrong even if it would prevent a nuclear apocalypse, a scenario the source article describes as highly unlikely. In addition to Tesla and SpaceX, Musk runs xAI, which competes directly with OpenAI and Google. A researcher who advises xAI recently developed a technique that could possibly alter the political leanings of large language models, as reported by WIRED.
The source article also notes that research on political bias in AI does not point in only one partisan direction. A study of Twitter’s recommendation algorithm published in 2021 showed that users were more likely to be shown right-leaning perspectives on the platform.
What remains uncertain
The practical effect of the new wording is still unclear. The source article says it is uncertain whether David Sacks, named by Trump in December as the White House AI and crypto czar, or anyone from the White House was involved in setting the new research agenda. It is also uncertain whether the new wording will substantially change the work most researchers are already doing.
At the same time, the political environment around NIST has changed. Since January, Musk’s so-called Department of Government Efficiency (DOGE) has been sweeping through the US government, effectively firing civil servants, pausing spending and creating an environment thought to be hostile to people who might oppose the Trump administration’s aims. DOGE has also targeted NIST, AISI’s parent organization, in recent weeks, and dozens of employees have been fired.
The White House did not immediately respond to WIRED’s request for comment. For now, the clearest change is in the official language: safety, fairness, misinformation and responsibility are less visible, while ideological bias and American competitiveness are more prominent.