Homeland Security needs far better oversight before it embraces AI

Recently, the U.S. Department of Homeland Security described itself as “early and aggressive adopters” of artificial intelligence tools. In October, Secretary of Homeland Security Alejandro Mayorkas issued principles for the department’s use of AI and named its first-ever chief AI officer. 

Given DHS’s 20-year track record, these moves are not reassuring. The agency has repeatedly used technology in ways that trample on the rights of millions of Americans. AI tools risk supercharging these harmful practices, and right now the department lacks oversight and accountability mechanisms strong enough to stop such abuses. Strengthening oversight would protect Americans from government abuse without making us less safe. DHS should make changes today.

A new report from the Brennan Center for Justice details the myriad ways in which the department’s oversight offices have consistently failed to check abuses. 

DHS has engaged in racial and religious profiling, surveillance of Americans for their political beliefs, many intelligence activities targeting journalists, the use of biased watchlists comprised overwhelmingly of Muslims, support for seemingly politicized prosecutions, broad sharing of junk intelligence, empowerment of police crackdowns, and the deployment of sprawling, pervasive surveillance tools along the U.S. borders. 

Yet in these and other instances, DHS oversight offices failed to fulfill their basic role of protecting civil rights and privacy, holding department decision-makers to account or providing redress to people impacted by aggressive agency actions.

Part of the problem is structural. Internal oversight is scattered among a patchwork of uncoordinated and poorly resourced offices, with little backing from department leadership. At Homeland Security headquarters, those offices include a civil rights and civil liberties office, a privacy office and an inspector general. Yet the civil rights and privacy offices are poorly staffed, with tiny rosters handling oversight of the department’s 260,000-plus workforce. 

And the civil rights and privacy offices lack strong representation in field offices, such as Border Patrol sectors where serious rights violations regularly occur. They also lack basic investigative powers, such as the ability to demand documents from the offices it oversees, and have little political clout and presence throughout the department. Civil rights and privacy representatives have at times been cut out of oversight of domestic intelligence where their input would be vital — as they were during the 2020 spying campaigns targeting racial justice demonstrators and journalists in the wake of George Floyd’s murder.

Most — though not all — of the department’s 22 component agencies, which include the Transportation Security Administration and Customs and Border Protection, have their own oversight offices. But these tend to focus on internal affairs — labor issues, promoting diversity, background checks and the like — rather than operations that directly affect the public, such as new facial recognition technology and algorithmic risk assessments. And they report not to headquarters’ oversight offices but to their agency’s leadership, setting up a clear conflict of interest and a competitive dynamic with headquarters oversight mechanisms. 

Indeed, it’s not clear that any neutral observers within the department writ large oversee much of what happens in these powerful agencies.

The DHS inspector general has more autonomy than the civil rights and privacy offices but spends a lot of its resources looking into bureaucratic or fiscal misdeeds, rather than the powerful DHS programs and systems that impact the public. And while it is charged with investigating civil rights complaints, it reviews few such complaints annually, out of hundreds received. The current inspector general is tarnished by a history of self-dealing and dysfunction, yet he remains in office, undermining its effectiveness and trustworthiness.

Taken together, these shortcomings have enabled repeated abuses of American’ rights with little accountability, as the Brennan Center has shown in a series of six reports.

That the Department of Homeland Security’s oversight structure is long overdue for top-to-bottom reform was clear before the department’s emphasis on AI. But change becomes even more urgent as the department adopts technologies that could massively amplify its impact on the lives of Americans by automating processes to identify individuals who constitute a security risk, capturing the faces and other biometrics of millions of people and deploying AI-powered tools to monitor social media.

Effective oversight at DHS is possible. Congress should strengthen the civil rights office’s standing and powers by granting it meaningful investigative authority and requiring that it publicize its assessments of DHS impacts on rights and liberties. Legislation was introduced in prior Congresses but went nowhere; it must be taken up again. Congress should also create an independent, robust office within DHS specifically to oversee the department’s vast intelligence operations.

The secretary of homeland security should install headquarters oversight personnel in agencies like Customs and Border Protection and their field operations, and promote training programs to strengthen core competencies, such as those related to new technology. He should require agencies to collaborate with their headquarters counterparts and ensure that oversight personnel actually assess and mitigate the impact of DHS’s advanced technology on Americans.

When reviewing DHS programs, the inspector general should more directly assess how these programs impact civil rights and privacy and educate the public better about what recommendations the department actually implements, and which they disregard.

The agency’s leadership has promised “extensive testing and oversight” of artificial intelligence tools, but DHS history does not inspire confidence. Only two months ago, the Government Accountability Office found that the department has been using facial recognition technology for years, in many cases without requiring staff to undergo training or robust policies to guard against abuses and transparency or accountability.

None of these reforms will impair Americans’ safety and will lead to fewer harmful and ineffective government programs. Congress and DHS’s secretary must act to strengthen oversight at the agency before the department’s headlong embrace of AI powers another two decades of abuse.

Spencer Reynolds is senior counsel in the Brennan Center for Justice’s Liberty and National Security Program and a former senior intelligence counsel in the Office of the General Counsel of the U.S. Department of Homeland Security.