- OpenAI is moving its government AI models under a new umbrella
- OpenAI for Government will provide AI at the federal, state, and local level
- The AI developer has already signed a pioneering deal with the DoD
OpenAI is consolidating its US government AI tools, such as ChatGPT Gov, under a single umbrella – OpenAI for Government.
OpenAI, alongside the likes of Anthropic and Meta, has partnered multiple times with the US government to develop new AI tools specialised for government workloads.
The new initiative will provide federal, state, and local governments with access to OpenAI’s most secure and compliant models, models specialized for national security, insight into upcoming models and tools, and support.
More AI models for government
Kicking off the new project, OpenAI has signed a $200 million with the U.S. Department of Defense’s Chief Digital and Artificial Intelligence Office (CDAO).
The project will focus on improving administrative work such as service member healthcare access, to boosting cyber defenses and data acquisition.
“Across these efforts, we’re aiming to improve both the day-to-day experience of public service and to help government employees feel more empowered, more efficient, and more supported in their critical missions,” OpenAI said.
“We are already seeing how OpenAI can help public servants at the state level spend less time on repetitive tasks and more time on high-impact work,” the announcement continued, with Open AI referencing the effectiveness of ChatGPT use for the Commonwealth of Pennsylvania, which saved employees around 105 minutes per day.
OpenAI is also deploying AI models at Los Alamos, Lawrence Livermore, and Sandia National Labs to improve scientific research, innovation, and national security.
“We are just getting started, and we look forward to helping U.S. government leaders harness AI to better serve the public. We are committed to working in close partnership with agencies, advancing their missions with powerful tools that are safe, and secure,” the company added.
