WASHINGTON — United Launch Alliance is piloting an early version of OpenAI’s government-compliant artificial intelligence chatbot, marking one of the first deployments of the technology designed specifically for defense contractors handling sensitive data.
The rocket manufacturer, jointly owned by Boeing and Lockheed Martin, has deployed what it calls “RocketGPT” to about 150 employees as part of a trial program. The system operates on Microsoft Corp.’s Azure secure cloud platform, which is approved for government data that must comply with International Traffic in Arms Regulations, or ITAR — strict security standards governing sensitive aerospace and defense information.
“We’re super excited,” ULA Chief Executive Tory Bruno told SpaceNews, describing the chatbot as a tool to help with “drudgery” and “tedious, time-consuming things” required for writing reports, drafting government proposals and analyzing flight telemetry, for example.
The deployment marks a step forward for AI adoption in the defense sector, where standard consumer versions of ChatGPT are prohibited due to security requirements.
Defense contractors must use AI tools built on government-authorized platforms, creating a specialized market that OpenAI has been cultivating.
Security compliance a priority
ULA, which builds and launches orbital rockets for the U.S. military and commercial customers, spent months working with OpenAI and consulting firm PwC to customize the chatbot for its specific needs. About 20 people from the three organizations collaborated on training and security compliance modifications.
Bruno emphasized that the AI system functions as a research assistant rather than a replacement for human workers, addressing concerns within the company about job displacement. “AI is really, really good at handling massive volumes of data,” he said. “So if that isn’t what you’re doing, that’s probably not the right tool.”
The ULA contract represents a significant victory for OpenAI, which has been actively pursuing government clients. The company recently secured a major Pentagon contract worth $200 million over one year to develop AI capabilities for national security applications.
The Defense Department’s chief digital and artificial intelligence office announced the contract on June 16, describing it as an effort to “develop prototype frontier AI capabilities to address critical national security challenges in both warfighting and enterprise domains.”
Measured expectations
Despite enthusiasm for the technology, Bruno cautioned against unrealistic expectations about AI’s capabilities. He criticized the narrative that companies can simply “sprinkle” AI on products to improve them automatically.
“You don’t just grab an AI off the shelf and just turn it on,” he said. “You have to train it on massive quantities of data.”
Bruno acknowledged that AI systems still make mistakes requiring human oversight, meaning employees remain accountable for final work products. The goal, he said, is to make workers faster and help them “generate higher quality products” rather than eliminate positions.
While Bruno declined to disclose the investment amount, he characterized it as “relatively modest” compared with expected productivity gains. ULA evaluated other AI platforms before selecting OpenAI based on what it viewed as superior large language model capabilities.
The pilot program includes employees across multiple departments, from engineering teams to legal and finance staff, as ULA explores the technology’s potential applications across its operations.