AI in Government: A Question of Trust

Written by Emily Wolfteich
Senior Industry Analyst

In May 2025, GovExec Intelligence asked the question: how much do you trust the AI you work with? After all, AI adoption has been a federal priority since 2019, first enacted under President Trump’s first term, and expanded during President Biden’s administration, where reported use cases across the federal civilian sector tripled from 710 in 2023 to 2,133 in 2024. New guidance from the Trump administration aims to expand AI procurement and use cases, including through data sharing and removal of regulations. In March of 2025, GovExec Intelligence’s survey found that 71% of respondents said that AI was more important to their agency than it was three months prior.

Across the government, AI is expected to be a mission enabler, a powerful tool, and a national security imperative. Yet for AI systems to be fully adopted, they need buy-in – and this is proving to be a challenge.

Our research finds that less than a third of federal respondents completely or mostly trust the AI system they work with, and 11% do not trust their AI systems at all. The majority (60%) “somewhat” trust their system.

Overall, the data indicates hesitancy. (“Somewhat” is not exactly a ringing endorsement.) This is not unexpected – five years is a relatively short time for paradigm-shifting technologies to be adopted, and AI systems have grown exponentially during that period. For federal agencies who still struggle with legacy technology equipment and face budgetary and regulatory constraints, it may be difficult to fully lean into the AI revolution.

Factors Affecting Trust

The research suggests a few possible contributing factors for this low level of trust.

General hesitancy among federal agencies to share data.

In February 2025, GovExec Intelligence found that 61% of respondents are somewhat or very hesitant to share data with other agencies, and only 4% are not at all hesitant. As new federal guidance pushes agencies to share both data and AI models, respondents who were already reluctant to share their data may need more clarity and reassurance that their data will be used securely and appropriately.

A.I. fatigue.

GovExec Intelligence’s February survey found that, when federal decision-makers were asked which work-related buzzwords they found most annoying, AI was at the top of the list. There has certainly been unprecedented chatter about artificial intelligence, including a slew of requirements for agencies that have changed from the previous administration, as well as a broader increase in companies looking to offer their AI solutions to the public sector, and non-governmental applications like ChatGPT that bring AI into the everyday.

AI errors.

Major problems with AI systems, particularly with generative AI, may be eroding federal confidence in its abilities. Generative hallucinations have hit big law firms across the country, leading to sanctions. Chatbots have encouraged businesses to break the law andlied to air passengers. While many of these errors have been more embarrassing than dangerous (a fake list of summer reading books, for example), and most of which would have been rectified by a quick quality check, there are also larger examples of workforce losses, defamation, and psychologicalimpacts. Particularly in federal agencies who deal with sensitive data, these public AI issues may decrease their trust in AI systems to securely handle their information.

Lack of public trust in AI.

These low trust levels correlate with a broader public distrust of artificial intelligence – a distrust that is growing. Edelman’s March 2025 Trust Barometer survey found that public trust in AI has dropped from 50% to 32% since 2019, and the Pew Research Trust reports that half of U.S. adult respondents are more concerned than excited about the increased use of AI in daily life.

What's Next?

Both government decision-makers and industry partners should be taking these low trust levels seriously. Edelman’s survey reports that the US public has one of the lowest trust ratings globally, and 40 percentage points less likely than the Chinese public to trust AI (32% versus 72%). This poses national security concerns. At the same time, public trust in tech companies to “do what is right” is the second-lowest in the world at 63%, a 10 percentage-point drop from 2024. This does not indicate an environment that is ready to trust in an AI revolution.

Federal workers are not immune from these public sentiments, but they are also tasked with AI adoption within their agencies. Finding the line between encouraging and supporting AI use, and demanding it without addressing some of the root causes of the mistrust, will be increasingly crucial to both promote innovation and curiosity, as well as to stay at the forefront of global AI dominance.

^{_{To read additional thought leadership from Emily, connect with her on LinkedIn.}}

Photo source: Anela R/peopleimages.com via Adobe Stock.

7 Ways BD & Sales Teams Can Use Federal Personas for Competitive Advantage

Written by Susan Rose VP, Strategic Marketing & Advisory Services Photo source: Anela R/peopleimages.com via…

Photo source: rrodrickbeiler via Adobe Stock.

Deep Dive: Department of Treasury

A look inside Treasury’s 2026-era tech strategy: AI isn’t a standalone budget line — but ~$48.8 M funding for a centralized fraud-detection platform points to growing use of chatbots, generative AI pilots, taxpayer services, fraud monitoring, and data-driven automation under its IT-modernization efforts.

Photo source: BeblackDesign via Adobe Stock.

Decoding OMB Memorandums M-25-21 and M-25-22

Explains Office of Management and Budget (OMB) M‑25‑21 and M‑25‑22 — new federal‑AI directives that replace prior guidance, empower agencies to adopt AI faster, and streamline procurement, while aiming to balance innovation, governance and public trust.