CAMBRIDGE, Mass. (AP) — After tech companies retreated from workplace diversity, equity, and inclusion (DEI) programs, they now face renewed scrutiny over DEI efforts in AI products. The White House and a Republican-led Congress have shifted focus from addressing harmful algorithmic discrimination to targeting what they call “woke AI. ” Recent subpoenas from the House Judiciary Committee to Amazon, Google, Meta, Microsoft, OpenAI, and others seek investigation into past efforts aimed at promoting equity in AI development and reducing biased outputs. Meanwhile, the U. S. Commerce Department’s standards arm has removed references to AI fairness, safety, and “responsible AI” in a recent call for researcher collaboration. Instead, it now emphasizes reducing “ideological bias” to promote “human flourishing and economic competitiveness, ” according to an Associated Press-obtained document. Tech workers are accustomed to fluctuating Washington priorities, but this new shift worries experts like Harvard sociologist Ellis Monk. Previously, he helped Google improve AI’s treatment of diverse skin tones, addressing biases in computer vision that historically misrepresented Black and brown individuals. Google adopted Monk’s color scale, replacing an outdated one designed primarily for white dermatology patients, which received positive consumer feedback. While Monk believes his scale’s use in many products remains secure, he fears future funding and initiatives aimed at inclusivity may decline due to the current political climate and market pressures. Former President Trump cut hundreds of DEI-related science and tech grants, though his administration’s impact on commercial AI development is more indirect. Republican Rep.

Jim Jordan, head of the judiciary committee, is investigating whether the Biden administration coerced AI firms to censor lawful speech. Michael Kratsios, director of the White House Office of Science and Technology Policy, recently criticized Biden’s AI policies as promoting social division and redistribution under the banner of equity. Before Biden’s term, growing evidence highlighted AI bias harms: studies found self-driving cars struggled to detect darker-skinned pedestrians; AI image generators overwhelmingly depicted surgeons as white men; facial recognition often erred with Asian faces; and misidentifications led to wrongful arrests of Black men. Google’s photo app infamously misclassified Black people as “gorillas, ” and even Trump administration scientists acknowledged uneven facial recognition performance by race, gender, and age. Biden’s election spurred increased industry focus on AI fairness. OpenAI’s 2022 ChatGPT launch accelerated AI application growth, pressuring companies like Google to catch up. However, Google’s Gemini AI chatbot rollout last year, intended to showcase inclusivity, instead fueled conservative backlash over “woke AI. ” Google’s image generator showed bias favoring lighter-skinned males and young women when depicting people in professions. Attempts to correct disparities led to overcompensation, such as depicting founding fathers as Black or Native American, prompting apologies and feature withdrawal, but also political controversy. At a February AI summit in Paris, Vice President J. D. Vance criticized “ahistorical social agendas through AI, ” referencing the Gemini incident, and vowed the Trump administration would ensure AI is free from ideological bias and protects free speech. Alondra Nelson, a former Biden science adviser and OSTP acting director, views the current focus on ideological bias as a recognition of longstanding concerns over algorithmic bias affecting housing, healthcare, and other areas. However, she doubts collaboration is possible given the political polarization that casts algorithmic and ideological bias as separate, conflicting issues.