US Government Sets New AI Safeguards to Enhance Transparency and Protect Rights

The White House has announced a significant policy directive concerning the use of Artificial Intelligence (AI) within federal agencies, aiming to set a new standard in safeguarding American rights and safety. By December 1, federal agencies utilizing AI are expected to enforce "concrete safeguards" as announced on Thursday. These measures come with the increasing integration of AI across a spectrum of governmental functions.

The directive originates from the Office of Management and Budget (OMB), ordering agencies to conduct meticulous monitoring and testing of the impact of AI technologies. This includes reducing the risks associated with algorithmic biases and providing the public with a transparent view of AI applications within the government sphere.

Agencies are mandated to undertake thorough risk assessments and establish precise operational and governance metrics. The focus is to create a protocol where AI usage, potentially impacting Americans' rights or safety, is coupled with publicly accessible, detailed disclosures.

In October, President Joe Biden signed an executive order under the Defense Production Act, stipulating that AI systems which could pose threats to national security or public wellbeing must undergo safety assessments. The results from such tests are to be shared with the US government prior to their public release.

This push for AI safeguards is detailed further by the White House's latest statement, highlighting specific areas such as the Transportation Security Administration's (TSA) use of facial recognition. New measures will enable air travelers to choose not to use this AI application, without hampering their screening process. In addition, federal healthcare AI systems, which aid in diagnostic decisions, are to be managed with human oversight for verification of outcomes.

Given the burgeoning development of generative AI, capable of synthesizing text, images, and videos from general prompts, the regulations aim to temper both excitement and concern. Potential risks identified range from employment disruptions, influence on election outcomes, to overarching fears of AI surpassing human control with destructive consequences.

The directive also necessitates governmental agencies to publish AI usage inventories, as well as reporting metrics on AI applications. Further, any AI code, models, or data owned by the government should be made public unless it imposes risks.

The directive comes against a backdrop of current AI applications by federal bodies. The Federal Emergency Management Agency (FEMA) uses AI in assessing hurricane damage, while the Centers for Disease Control and Prevention (CDC) employ AI for predicting disease patterns and identifying opioid consumption. Moreover, the Federal Aviation Administration (FAA) leverages AI to manage air traffic, consequently enhancing travel times in busy metropolitan areas.

To support these endeavors, the White House aims to recruit 100 AI professionals focused on promoting safe AI technologies. Additionally, within the next 60 days, federal agencies are to appoint chief AI officers, emphasizing the commitment to this technological trajectory.

Finally, within broader security concerns, the Biden administration in January proposed "know your customer" obligations on U.S. cloud service providers to prevent foreign entities from using US data centers for AI model training.