The Power of AI:
Artificial Intelligence (AI) is increasingly being integrated into many sectors of government, and records management is no exception. For instance, the National Archives and Records Administration (NARA) has tested AI to help with the classification and retrieval of digital records, aiming to improve efficiency by automating processes that are typically labor-intensive. AI tools are being used to identify relevant records, classify them into categories, and even flag documents for potential archival value, eliminating the need for human review in some cases.
Another example is the U.S. Department of Veterans Affairs (VA), which has piloted AI-powered tools to help manage veteran records more efficiently, cutting down on processing times for benefits claims. Meanwhile, the Department of Defense (DoD) has tested auto-filling record metadata with AI to reduce unnecessary manual labor.
In short, AI holds the potential to handle the sheer volume and complexity of government records in ways that humans simply cannot. AI can rapidly analyze thousands of documents, pulling out key information and identifying patterns that would take people much longer to detect. These early adopters are showing how AI could transform government records management by making it faster, more accurate, and more informationally fruitful.
The Limits of AI:
However, as promising as AI is, governments are already encountering challenges. One prominent example involves OpenAI’s new contract with the U.S. Department of Homeland Security (DHS). OpenAI altered its terms of service in a way that raises important questions about whether the output generated by AI could legally be considered a federal record and, if so, who has responsibility for it.
According to OpenAI, the answer is that it is fully the government agency’s responsibility: their terms now read, “For clarity, any Federal Records-related obligations are Agency’s, not Company’s.”
For what it’s worth, this kind of distinction may be an argument in favor of function-specific platforms that incorporate AI capabilities rather than AI-specific platforms that can be used for a wide variety of applications. In other words, if your agency uses an electronic records management (ERM) platform with AI features, any AI output would already be contained within the system. However, if your agency uses a third-party AI tool to generate output that might count as a federal record, that output will have to be somehow captured and managed in a separate system according to a separate set of processes.
The Uncertainties of AI:
This brings up a broader question of ownership and accountability: if an AI system generates a document, who is responsible for maintaining and ensuring the integrity of that record? “The question of who bears legal and moral responsibility for harms caused by ML AIs will … be a difficult issue to settle going forward,” writes ARMA (formerly the Association of Records Managers and Administrators) Magazine.
In fact, ARMA points out that, in some cases, records of AI-driven decision-making processes may need to be captured…but how? Consider FOIA requests. Multiple government entities, including both the State Department and the National Archives and Records Administration, are testing use of AI in responding to FOIA requests. But if AI is being used, is its role being noted? “When FOIA officers conduct a search and review of records responsive to a FOIA request, there generally need to be notes on how the request was processed, including, for example, the files and databases the officer searched for records,” Beryl Lipton, an investigative researcher at the Electronic Frontier Foundation, tells FedScoop. “If AI is being used in some of these processes, then this is important to cover in the processing notes, because requesters are entitled to a search and review conducted with integrity.”
The Need for People:
Despite AI’s growing role, human oversight remains essential in records management. People are still better at making nuanced decisions about what records are significant and require preservation versus those that do not. For instance, records management involves compliance with legal requirements, which often necessitates a level of human interpretation that AI cannot yet match.
Moreover, as noted above, AI lacks the ability to explain its decision-making processes clearly, which can create challenges in ensuring transparency and accountability—core principles in government records management. This underscores the need for human professionals to review and validate the work AI systems perform.
“While AI is making headway at improving data management, there are still many things that it can’t do,” writes the MIT Sloan Review in an analysis of how AI is improving data (and, we would add, records) management. The Sloan Review points out that AI isn’t yet capable of:
- Creating an effective records strategy,
- Affecting organizational culture around records,
- Developing appropriate governance policies and structures,
- Discerning whether the organization is making good use of its records or if it’s even capturing the right records or the wrong records,
- And more.
Those capabilities may well come one day, but for today, they represent critical functions only humans can execute.
Ultimately, the use of AI in government records management is poised to grow, offering both incredible potential and significant challenges. In the coming months, we will be delving deeper into these topics, driven by our own work to integrate AI into TransAccess GovCloud Records in ways that are equal parts effective, thoughtful, and conscientious. Our goal is to harness AI’s power while mitigating its risks, ensuring that our users can fully leverage its potential without compromising on security, compliance, or the human touch that remains so essential.
About PSL
PSL is a global outsource provider whose mission is to provide solutions that facilitate the movement of business-critical information between and among government agencies, business enterprises, and their partners. For more information, please visit or email info@penielsolutions.com.