GAO’s publish survey about facial recognition technology (FRT) activities

Facial recognition—a type of biometric technology—mimics how people identify or verify others by examining their faces. Recent advancements have increased the accuracy of automated FRT resulting in increased use across a range of applications. As the use of FRT continues to expand, it has become increasingly important to understand its use across the federal government in a comprehensive way.

GAO was asked to review the extent of FRT use across the federal government. This report identifies and describes (1) how agencies used FRT in fiscal year 2020, including any related research and development and interactions with non-federal entities, and (2) how agencies plan to expand their use of FRT through fiscal year 2023.

GAO surveyed the 24 agencies of the Chief Financial Officers Act of 1990, as amended, regarding their use of facial recognition technology. GAO also interviewed agency officials and reviewed documents, such as system descriptions, and information provided by agencies that reported using the technology.

Recent advancements in facial recognition technology have increased its accuracy and its usage. Our earlier work has included examinations of its use by federal law enforcement, at ports of entry, and in commercial settings.

For this report, the GOA surveyed 24 federal agencies about their use of this technology.

- 16 reported using it for digital access or cybersecurity, such as allowing employees to unlock agency smartphones with it
- 6 reported using it to generate leads in criminal investigations
- 5 reported using it for physical security, such as controlling access to a building or facility
- 10 said they planned to expand its use

In response to GAO's survey about facial recognition technology (FRT) activities in fiscal year 2020, 18 of the 24 surveyed agencies reported using an FRT system, for one or more purposes, including:

Digital access or cybersecurity. Sixteen agencies reported using FRT for digital access or cybersecurity purposes. Of these, 14 agencies authorized personnel to use FRT to unlock their agency-issued smartphones—the most common purpose of FRT reported. Two agencies also reported testing FRT to verify identities of persons accessing government websites.

Domestic law enforcement. Six agencies reported using FRT to generate leads in criminal investigations, such as identifying a person of interest, by comparing their image against mugshots. In some cases, agencies identify crime victims, such as exploited children, by using commercial systems that compare against publicly available images, such as from social media.

Physical security. Five agencies reported using FRT to monitor or surveil locations to determine if an individual is present, such as someone on a watchlist, or to control access to a building or facility. For example, an agency used it to monitor live video for persons on watchlists and to alert security personnel to these persons without needing to memorize them.

Ten agencies reported FRT-related research and development. For example, agencies reported researching FRT's ability to identify individuals wearing masks during the COVID-19 pandemic and to detect image manipulation.

Furthermore, ten agencies reported plans to expand their use of FRT through fiscal year 2023. For example, an agency plans to pilot the use of FRT to automate the identity verification process at airports for travelers.

Improved Performance Planning Could Strengthen Technology Transfer

A Department of Energy national lab developed a battery that now powers some hybrid and electric cars. But how do new energy technologies get from the lab to the market?
Transferring technologies from the DOE to private companies isn't always easy. Barriers such as the "valley of death"—a gap between the end of public funding and the start of private funding—can stop a transfer.
The Department of Energy (DOE) and its national labs have taken several steps to address potential barriers to technology transfer—the process of providing DOE technologies, knowledge, or expertise to other entities. GAO characterized these barriers as (1) gaps in funding, (2) legal and administrative barriers, and (3) lack of alignment between DOE research and industry needs. For example, the “valley of death” is a gap between the end of public funding and start of private-sector funding. DOE partly addresses this gap with its Technology Commercialization Fund, which provides grants of $100,000 to $1.5 million to DOE researchers to advance promising technologies with private-sector partners. Further, DOE's Energy I-Corps program trains researchers to commercialize new technologies and to identify industry needs and potential customers. However, DOE has not assessed how many and which types of researchers would benefit from such training. Without doing so, DOE will not have the information needed to ensure its training resources target the researchers who would benefit most.
DOE plans and tracks the performance of its technology transfer activities by setting strategic goals and objectives and annually collecting department-wide technology transfer measures, such as the number of patented inventions and licenses. However, the department does not have objective and measurable performance goals to assess progress toward the broader strategic goals and objectives it developed. For example, without a performance goal for the number of DOE researchers involved in technology transfer activities and a measure of such involvement, DOE cannot assess the extent to which it has met its objective to encourage national laboratory personnel to pursue technology transfer activities. Internal control standards for government agencies call for management to define objectives in measurable terms, either qualitative or quantitative, so that performance toward those objectives can be assessed. Moreover, DOE has not aligned the 79 existing measures that it collects with its goals and objectives, nor has it prioritized them. Some lab stakeholders said that collecting and reporting these measures is burdensome. Prior GAO work has found that having a large number of performance measures may risk creating a confusing excess of data that will obscure rather than clarify performance issues.