State Department conducting market research on an LLM it could customize

The Department of State is looking into off-the-shelf large language models that could be customized for its use cases and be used with an array of government data, according to a recent disclosure from the agency.

The sources sought notice, which was published last week, is part of the agency’s market research to figure out what tools are available for purchase. Specifically, the department is looking for a platform “to support a range of AI-enhanced workflows able to securely process and analyze government data of various classification levels.”

According to the notice, the LLM must meet the Defense Department’s Impact Level 6 and have at least a moderate-level FedRAMP authorization for commercial software. The publication isn’t a solicitation or indication that the agency will commit to a contract award. 

The State Department has already launched an internal AI chatbot, is encouraging people to use things like ChatGPT, and is using the technology to complement its diplomatic mission. In remarks in June, Secretary of State Antony Blinken said AI has freed up the department to focus on what employees can add.

There are also 28 uses of AI on the department’s public use case inventory, including deepfake detection and conflict forecasting tools.

In responses, the agency wants to hear from companies to provide information about products that can leverage generative AI in processing content; support “robust, secure translation capabilities;” have “mature data management capabilities;” and are “highly secure” with “comprehensive access controls,” among other things.

According to the notice, responses are due Aug. 16. The department didn’t immediately respond to a request for comment.