A confidential internal memo, signed by Acting ICE Director David Venturella and leaked to The Intercept by agency employees, reveals that U.S. Immigration and Customs Enforcement has quietly launched a “Doxing Mitigation Initiative” powered by artificial intelligence firm ZeroFox — a company with a documented history of surveilling political activists and categorizing dissidents as threat actors. The program, backed by a $15 million federal contract, is ostensibly designed to protect the identities of ICE personnel. But current agency officials who spoke anonymously to The Intercept are raising a different and more troubling possibility: that the same technology could be turned inward, against the agency’s own whistleblowers.
The Memo and the Contract
The internal communication, addressed to all ICE employees, describes ZeroFox as “a modernized protective technology solution designed to strengthen safeguards for our most valuable resource: You and your families.” The framing is protective, even paternal. But the contractual reality is more expansive. According to public procurement records reviewed by The Intercept, the Department of Homeland Security — ICE’s parent agency — entered a $15 million contract with ZeroFox in July for software licenses to support ICE operations and investigations. Those records indicate ZeroFox will provide services specifically for the Office of Intelligence at ICE’s Homeland Security Investigations division.
ZeroFox markets itself as an AI-driven platform that monitors social media and the dark web for data leaks and threats. “ZeroFox never stops finding, mapping, and monitoring your digital presence,” the company states on its website. What remains unclear from either the contract language or the internal memo is precisely how the Doxing Mitigation Initiative will function in practice. Neither ICE nor ZeroFox responded to The Intercept’s requests for comment.
A Company With a Complicated Past
ZeroFox is not a neutral actor in the surveillance landscape. The Intercept’s own prior reporting documents that ZeroFox technology was previously used to identify and categorize individuals engaged in political activism — including during the public unrest that followed the 2015 police killing of Freddie Gray in Baltimore and during the January 6, 2021 attack on the U.S. Capitol. The FBI has also contracted with ZeroFox; that earlier arrangement included the labeling of Black Lives Matter organizers as “threat actors” in the company’s monitoring systems.
That history matters when evaluating what ZeroFox’s AI tools are capable of doing — and what they may be configured to do when deployed inside a federal immigration enforcement agency. The platform is built to find, map, and track digital identities. The contract now places those capabilities within an intelligence office whose mandate includes undercover operations and the monitoring of individuals inside the United States.
“It Sounds Like Inside Surveillance”
The most striking reaction to the new initiative did not come from outside critics — it came from within ICE itself. An agency official, speaking to The Intercept on condition of anonymity due to fear of retaliation, described the program in blunt terms: “It sounds like inside surveillance.”
The official went further, articulating a specific fear about how ZeroFox’s tools could be leveraged against employees who speak out about agency conduct. “If things go sideways they may just throw all your info out there, make you out to be the problem or dox you,” the official said. “I don’t trust it.”
The concern is not hypothetical. The Trump administration has shown a sustained interest in identifying and neutralizing leakers and whistleblowers across the federal government. An AI platform capable of mapping the digital presence of individuals — deployed inside an intelligence office — creates an infrastructure that could as easily be pointed at dissenters within the agency as at external threats.
The Whistleblower Protection Gap
The risk flagged by the unnamed ICE official lands in an already precarious legal environment for government employees who choose to speak out. As the National Whistleblower Center has noted in its advocacy for the AI Whistleblower Protection Act — a bipartisan bill introduced by Senate Judiciary Chair Chuck Grassley in May 2025 — individuals who work within or alongside AI systems currently lack explicit statutory protections for reporting potential violations.
The proposed legislation would cover disclosures related to what it defines as an “AI security vulnerability” or an “AI violation,” the latter meaning any breach of federal law occurring during or related to the development, deployment, or use of artificial intelligence, or any inadequate response to a concrete risk that AI may pose to public safety, public health, or national security. The bill would prohibit employers from discharging, demoting, suspending, threatening, blacklisting, or harassing covered individuals in retaliation for protected disclosures.
But that bill has not yet passed. In its absence, federal employees who witness the misuse of AI-powered surveillance tools — including, potentially, tools like those ZeroFox provides — may have limited recourse. The gap between technological deployment and legislative protection is precisely the environment in which the Doxing Mitigation Initiative is being rolled out.
Dual-Use Technology and the Question of Accountability
Cybersecurity experts who study this space have consistently pointed to the dual-use nature of identity-monitoring platforms. A tool capable of suppressing an agent’s personal information from public databases is, by technical design, also a tool capable of surfacing that same information about anyone else — including colleagues who might become inconvenient to agency leadership.
The specific placement of ZeroFox’s contract within ICE’s Office of Intelligence at the Homeland Security Investigations division is significant. HSI has been the subject of recent scrutiny for its use of undercover agents in domestic settings, and its intelligence office operates with considerable latitude. Embedding a commercial AI surveillance platform into that structure — with a $15 million commitment from public funds and minimal public disclosure — represents a meaningful expansion of the agency’s digital monitoring infrastructure.
What the internal memo frames as a benefit for employees — protection from doxing, safeguarding families — is, from a technical standpoint, the deployment of a continuous, AI-driven identity surveillance apparatus inside a federal law enforcement agency. The protection flows in whatever direction the platform is configured to point.
Leaked From Within
It is worth noting that the memo itself became public because ICE employees chose to leak it. The agency’s own workforce, or at least some portion of it, viewed the program as significant enough — and concerning enough — to bring to journalists rather than accept in silence. That act of disclosure, performed at personal risk by individuals who fear retaliation, is precisely the kind of conduct the AI Whistleblower Protection Act is designed to protect.
For now, those employees have no explicit statutory shelter. The technology they are reporting on is already funded, already contracted, and already being rolled out under the authority of a $15 million federal agreement that received no public announcement.
What Remains Unknown
Critical questions remain unanswered. The operational specifics of the Doxing Mitigation Initiative — what data it collects, how it is stored, who has access, and what triggers a review of an individual’s digital profile — have not been disclosed. ICE has not clarified whether the initiative applies exclusively to external threats or whether it encompasses internal monitoring of agency personnel. ZeroFox has not described the technical architecture of the tools being provided under this specific contract.
The public procurement record confirms the contract’s existence and its value. The leaked memo confirms the program’s launch and its framing. What lies between those two documents — the actual mechanics of a $15 million AI surveillance deployment inside a federal immigration enforcement agency — remains, for now, hidden.
This article draws on reporting from The Intercept and legislative analysis from the National Whistleblower Center.
