The Biden administration is laying the foundation for greater use of artificial intelligence tools across the federal government — but agencies are already charting a path for how federal employees will work with AI tools.
The State Department is relying on AI tools to help its workforce go through a massive inventory of diplomatic cables and recommend documents to declassify. The department also plans to soon release its first enterprise AI strategy.
Giorleny Altamirano Rayo, the State Department’s chief data scientist and responsible AI official, said the upcoming strategy will ensure trustworthy AI is part of the department’s focus on building up its data analytics capabilities.
The State Department’s Center for Analytics, led by its Chief Data Officer Matthew Graviss, is developing the enterprise AI strategy.
“That’s going to guide our implementation of AI in a way that’s responsible, and that is going to keep the trust in our department on using AI,” Rayo added.
Rayo said the upcoming AI strategy is closely linked to the department’s first enterprise data strategy, as well as Secretary of State Antony Blinken’s broader plan to modernize U.S. diplomacy.
“We basically consider it another tool in our diplomatic toolbox. But we’re very mindful of this critical connection between data quality and successful use of AI,” Rayo said in a July 12 FedInsider webinar.
Rayo said the upcoming strategy will focus on how the department can field AI tools in a way that upholds privacy, security, ethics and equity to make these applications “trustworthy for the end-user.”
“Responsibility is the name of the game, and that responsibility is paramount in applications of AI. We build an audit, and we scrutinize our models, so we make sure that we’re doing it the right way,” Rayo said. “We’re doing all of this by being mindful of the importance of both policy, governance and strategies that allow us to do this responsibly.”
In the lead-up to the AI strategy’s release, the Center for Analytics recently wrote a new chapter in the Foreign Affairs manual on the responsible use of AI. That chapter sets out the department’s principles for using AI, and contains an inventory of AI use cases.
Rayo said the AI language in the Foreign Affairs manual echoes the Trump administration’s December 2020 executive order on how to design, develop and deploy and use AI in a responsible manner across the federal government.
“The first thing that we really wanted to do is make sure that we were following directives and that we’re following the values of equity, privacy and respect for civil liberties,” Rayo said.
The State Department is also using AI to assist in its declassification work.
Rayo said the department is using machine learning to augment what its workforce is doing to declassify diplomatic cables.
Diplomatic cables at the State Department remain classified for 25 years. At that point, the department reviews the documents to determine whether where they should remain classified or can be fully or partially declassified.
“There was a huge explosion of documents [and] emails around the world. And the State Department just got bombarded with tens of millions and thousands of millions of emails to sift through,” Rayo said about the latest tranche of records that are scheduled for review.
The department is using natural language processing, to review these documents.
Rayo said those AI tools are reaching the same declassification decision as human reviewers about 97% of the time, and have reduced the time to review these records by about 65%.