Has the tech industry made its large language models (LLMs) for the government “anti-woke,” in accordance with the Trump administration’s policies? Not yet, tech giants say.
While companies wrestle with issues around “wokeness,” they’re offering federal agencies significant discounts on their core AI products, according to a slew of recent announcements by the federal government.
Tech companies offering chatbots and other AI tools, including Google, OpenAI, and Box, say negotiations for federal contracts do not yet involve conversations about the “anti-woke” executive order taking aim at perceived diversity, equity, and inclusion (DEI) in the development of LLMs.
Josh Gruenbaum, the Federal Acquisition Service (FAS) commissioner for the General Services Administration, told IT Brew it wouldn’t “make sense” for negotiations with these companies to include navigating the order, as the government has not yet released compliance guidance to accompany the order.
Gruenbaum said, however, “the concept of an executive order and how we are going to prosecute now, being in privity of contract with these frontier models, [the makers of the models] know that these are our expectations going forward.”
The FAS director also claimed the discounts offered to the government at large are a “real acknowledgement from industry since January 20, 2025, that President Trump and the administration that he has built around him is an administration that is very focused on common sense, and particularly with a strong business sense.”
Financial Times reported in July, conversely, that a leading technology group who spoke to the outlet described dealing with the Trump administration as a “circus,” while an unnamed government contractor described the administration as “super aggressive” and having “upset CEOs and others.”
Waiting for a signal. Joseph Larson, OpenAI’s VP and head of government, said that the company has not yet seen the guidance from the government, but believes the company is consistent with the anti-woke executive order on AI. “We’re waiting for the government to tell us how to certify that in terms of compliance.”
Larson acknowledged that once the government releases guidance, the company will address questions directly to the government “after we’ve absorbed that.”
When asked about whether Gemini for Government meets the Trump administration’s EO on “ideologically neutral models,” Brent Mitchell, VP of go-to-market for Google Public Sector, said that the platform is a public-sector specific instance (i.e., a dedicated and customized AI buildout) and the cloud environment is “100% controlled by our customers.”
According to Mitchell, Google’s models aren’t trained on customer data “whatsoever.”
Principle changes. As Google wrote in its 2023 AI Principles Progress Update, a key part of the company’s machine learning (ML) work “involves developing techniques to build models that are more inclusive.”
Top insights for IT pros
From cybersecurity and big data to cloud computing, IT Brew covers the latest trends shaping business tech in our 4x weekly newsletter, virtual events with industry experts, and digital guides.
At the time, the company stated it had strengthened its community-based research efforts with a focus on “historically marginalized communities or groups of people who may experience unfair outcomes of AI.” The efforts include evaluations of gender-inclusive health to scaling speech recognition to recognize “underrepresented languages and adapt to new languages and data.”
However, TechCrunch reported in March that Google had removed the mentions of diversity and equity from its Responsible AI web page.
On that page, the company states that foundational ML research methodologies aim to represent AI at its best, which it defines as: “responsible, transparent, robust, and inclusive.” It also states the team focuses on fairness through identifying and preventing “unjust prejudicial treatment of people, when and where they manifest in algorithmic systems.”
Mitchell told IT Brew in an email the company is committed to responsible AI practices in building safety filters, fairness checks, and updating its models consistently.
“As model providers we enable government partners to validate the accuracy and help guide tone and response,” Mitchell wrote. “Because generative AI can reflect real-world biases, we’ve developed open-source tools and datasets to help identify and reduce bias, and we partner with external researchers to advance this work across academia and industry.”
The clearance rack. As well as giving a 71% discount to the government for its workspace suite, Google is offering its Gemini chatbot for $0.47 for each federal agency. Gemini Business is typically priced at $20, which makes the discount for the tool 97.65%.
Ann Marie Clements, the director of communications for Google Public Sector, confirmed to IT Brew that the discounted offering of Gemini for Government priced at $0.47 is related to President Trump being the 47th US president.
Gruenbaum confirmed this as well.
“I think it’s really cool, right?” Gruenbaum said. “I thought it was really interesting, and a real hat-tip to…my point where industry is recognizing that this is a historic administration with a historic president and they want to be a part of that.”
FT reported that Gruenbaum’s “sometimes brusque approach—demanding that the likes of Microsoft, Amazon, and Uber identify savings or face the possible cancellation of lucrative contracts” has damaged relationships with the tech and consultancy industries. Gruenbaum disputed these claims to the outlet, saying all negotiations had happened in good faith.
Box CEO Aaron Levie told IT Brew the company was not asked to incorporate any kind of reference to the presidency, including a 47, into any of the pricing options.