The National Institute of Standards and Technology (NIST) has recently revised its instructions for scientists collaborating with the US Artificial Intelligence Safety Institute (AISI). This update eliminates references to “AI safety,” “responsible AI,” and “AI fairness” in favor of a priority on “reducing ideological bias” to promote human flourishing and economic competitiveness. This shift is introduced in a new cooperative research agreement sent out in early March, which contrasts with previous directives urging researchers to address discriminatory behaviors in AI models related to gender, race, age, and socioeconomic status.
The updated agreement also removes expectations for developing tools aimed at authenticating content and tracking misinformation, suggesting a decreased focus on combating misinformation and deep fakes. Furthermore, it emphasizes a push for American dominance in AI, instructing one group to create testing tools to enhance the nation’s global standing in AI technology.
Critics, including an anonymous researcher associated with the AI Safety Institute, express concern that the removal of safety and fairness priorities could expose everyday users to unfair and discriminatory AI systems, ultimately worsening outcomes for those who are not affluent. "If you’re not a tech billionaire, this is going to lead to a worse future for you and the people you care about," the researcher states.
Another researcher previously involved with the AISI questioned the implications of such changes, asking, “What does it even mean for humans to flourish?”
Elon Musk, who is now taking steps to cut government bureaucracy under President Trump, has lampooned AI models from companies like OpenAI and Google, labeling them as “racist” and “woke.” His critiques include incidents where algorithms have wrestled with nuanced ethical dilemmas. Musk’s involvement in AI through his own company, xAI, further positions him at the center of the discourse around AI biases.
Research suggests that political biases in AI can affect perspectives across the spectrum, as demonstrated in a 2021 analysis of Twitter’s algorithm that favored right-leaning viewpoints. Since January, Musk’s Department of Government Efficiency (DOGE) has affected numerous government positions, fostering an environment often viewed as antagonistic toward dissent against the Trump administration’s stances. At NIST, dozens of employees have reportedly been dismissed.
Stella Biderman, executive director of Eleuther, a nonprofit engaged with the AISI, indicated that the recent policy changes are directly influenced by the Trump administration’s objectives. She added, "It isn’t surprising to me that rewriting the plan was necessary to continue to exist."
Since taking office, Trump has rescinded Biden’s executive order that established the AI Safety Institute—initially aimed at addressing the potential threats of AI, such as misuse in cyber warfare or biosecurity. The new directive focuses on adopting AI systems free from political bias or preconceived social agendas, as stated in a recent executive order.
At the AI Action Summit in Paris, Vice President JD Vance affirmed the administration’s commitment to ensuring American competitiveness in AI development, downplaying the need to be overly concerned about AI safety.
The changes have raised alarms among researchers, who are worried that their engagement with political figures may inadvertently lead them to overlook the consequences of policy shifts that prioritize power over equitable AI practices. Despite the uncertainty around these revised policies, the potential consequences for AI governance and public accessibility remain significant.
