Now US government agencies can use OpenAI’s ChatGPT too

Everyone wants to get their hands on AI tools, and employees in government agencies are no exception. Still, stakes are far higher for governments when it comes to security, privacy, and protecting the public.

As a means of overcoming these challenges, OpenAI has rolled out ChatGPT Gov, a version of its flagship frontier model specifically tailored to US government agencies. The platform has many of the same capabilities as OpenAI’s other enterprise products, including access to GPT-4o and the ability to build custom GPTs — and it also features a much higher level of security than ChatGPT Enterprise.

The launch comes after a busy week for OpenAI (and the industry in general): US President Donald Trump has rolled back what he called “barriers” to building and launching AI, top players have partnered on the ambitious “Stargate Project,” and open-source newcomer DeepSeek-R1 has emerged to potentially disrupt the industry.

“We believe the US government’s adoption of artificial intelligence can boost efficiency and productivity and is crucial for maintaining and enhancing America’s global leadership⁠ in this technology,” OpenAI wrote in a blog post. “We see enormous potential for these tools to support the public sector in tackling complex challenges — from improving public health and infrastructure to strengthening national security.”

AI in the spotlight

AI is almost always in the spotlight, but this past week has significantly ratcheted up interest in the technology.

Notably, on January 21, OpenAI announced Project Stargate, its new partnership with Microsoft, Oracle, Arm, Nvidia, SoftBank and MGX. The initiative is pledging $500 billion to build AI infrastructure, $100 billion of which will be invested “immediately.”

Following that — and just four days into his presidency — Trump announced an executive order that revokes some existing AI policies and directives that he said could potentially inhibit AI innovation in America. This clears a path for the US to “act decisively to retain global leadership in artificial intelligence,” the order said.

And now OpenAI and other major players face a new, much less expensive open-source contender: DeepSeek-R1. The model has been the talk of the tech world, as it purportedly matches o1’s performance, is 90% to 95% cheaper, and supposedly much faster than o1.

In its first week of launch, DeepSeek quickly surged past competitors as the most downloaded model on both Hugging Face and Apple’s App Store.

However, the Chinese-made app’s meteoric rise is concerning for some, particularly as the US pumps more and more money into its AI efforts with an aim to become the leader in the technology. Trump said it should be seen as a “wake-up call,” even as he praised DeepSeek as “very much a positive development” for AI overall, because it means enterprises don’t have to spend billions on AI tools that will give them a competitive edge.

Government moving at ‘atypical’ pace

Governments around the world are compliance-heavy and regulation-guided, and have a “certain sort of lag” around how they offer services, Alison Brooks, research VP for smart cities and communities — public safety at IDC, told InfoWorld. They also continue to struggle with employee recruitment and retention, and other constraints.

“The government is really looking at this tech, they just recognize the ability for these systems to augment their resources,” she said. “They are approaching as a fast follower — that’s atypical of government.”

There are numerous use cases across government, particularly in public safety, she noted. For instance, amidst emergencies like the California wildfires, agencies could create consolidated alerts and guidelines and “put those out there in multiple different languages immediately,” said Brooks.

Or, in procurement, a request for proposals (RFP) used within one city government could be repurposed by another city agency with just a few simple updates.

Brooks also pointed to a use case of a joint investigation team (JIT) of French and Dutch authorities. They cracked encrypted messages in an international criminal communications network known as the Matrix that contained more than 2 million encrypted messages in 33 languages, spanning 40 servers. Generative AI helped to instantly translate messages and prioritize those that were most likely to cause injury to life or property — what otherwise would have been a “logistical struggle,” she noted.

With tools like ChatGPT Gov and partnerships with OpenAI, government agencies “have more of a feeling that they can do these quasi-experimental innovative things in a way that isn’t going to get them in the papers,” said Brooks.

Getting GPT into the hands of government users

OpenAI noted that more than 90,000 users across 3,500 federal, state, and local governments in the US have been using ChatGPT since 2024, sending more than 18 million messages on the platform, and the new ChatGPT Gov meets increasing demand in government.

The platform provides access to OpenAI’s flagship model, GPT-4o, which processes text and images, produces summaries, and performs coding and math. Government employees will be able to build and share custom GPTs, upload text and image files, and save and share conversations. The platform also features an administrative console for CIOs and IT teams to manage users, groups, custom models, and single sign-on (SSO) functions.

Agencies can deploy ChatGPT Gov in their Microsoft Azure commercial cloud or Azure Government cloud, on top of Microsoft’s Azure’s OpenAI ⁠service. According to OpenAI, the self-hosted model will allow government agencies to more easily manage security, privacy, and compliance requirements, including cybersecurity frameworks such as FedRAMP High (for highly sensitive, classified or mission-critical information), IL5 (for controlled unclassified information), and International Traffic in Arms Regulations (for the export of defense and military technologies).

The latter seems to indicate that ChatGPT Gov has a higher level of security than OpenAI’s other products; in fact, the company noted that it is working toward FedRAMP Moderate and High accreditations for its fully-managed SaaS ChatGPT Enterprise platform. 

“OpenAI’s ChatGPT has made significant progress supporting agency mission requirements since last year, when many agencies banned usage and labeled it as an ‘unnecessary risk,’” Daniel Snyder, Gartner senior director analyst and KI (key initiative) leader, told InfoWorld.

He pointed out that many initial concerns centered around data security risks in ChatGPT. This launch “will likely reduce many data concerns and result in wider usage as agencies seek to experiment with OpenAI’s frontier models,” he said.

How the Air Force and several states are already using ChatGPT Enterprise

OpenAI has been working with early adopter government agencies for some time now. For instance, the Air Force Research Laboratory is using ChatGPT Enterprise to help improve access to internal resources, support AI education efforts, and to perform basic coding tasks.

Meanwhile, Commonwealth of Pennsylvania⁠ employees are participating in a pilot program experimenting with ChatGPT Enterprise. So far, the model has helped reduce time spent on routine tasks like analyzing project requirements by roughly 105 minutes per day, according to OpenAI. The State of Minnesota’s Enterprise Translations Office is also using ChatGPT to deliver faster, more accurate translation services to multilingual communities, “significantly reducing costs and turnaround times.”

“We look forward to collaborating with government agencies to enhance service delivery to the American people through AI and to foster public trust in this critical technology,” OpenAI said.

By making AI available to the US government, the company said it will “ensure AI serves the national interest and the public good, aligned with democratic values, while empowering policymakers to responsibly integrate these capabilities to deliver better services to the American people.”