Department of Homeland Security and Google partner on $1.5M contest for algorithm that identifies concealed items in images captured by body scanners
Context & Ripple Effects
This 2017 contest is an early data point in what became a durable working relationship between DHS and the commercial AI industry: the agency later stood up a dedicated AI task force to explore uses from cargo screening onward, and by 2024 was running pilot programs with OpenAI, Anthropic, and Meta against trafficking and drug crimes. The $1.5M prize structure — outsourcing detection R&D to outside teams rather than building in-house — is the procurement model those later partnerships scaled up.
The pairing also carries a known civil-liberties overhang. DHS's biometric ambitions have repeatedly drawn pushback: it dropped its proposal to scan US citizens at borders after criticism, and the ACLU sued to force disclosure of its airport face-recognition program. A Google-built screening algorithm sits squarely in that contested zone, especially given reporting on a Google contractor's dubious collection of facial scans from people of color.
First-order effects
- Competing teams now chase a $1.5M purse by building classifiers that flag concealed items in body-scanner images, giving DHS a pipeline of candidate detection algorithms without funding internal development.
- Google converts the contest into a standing credential as a homeland-security AI vendor, positioning it ahead of rivals for follow-on federal screening work.
Second-order effects
- Other frontier labs take note of the template: within seven years DHS formalizes the arrangement into named partnerships with OpenAI, Anthropic, and Meta, turning ad-hoc prizes into recurring pilot programs.
- Civil-society groups gain a concrete target — any deployed concealed-item detector joins the same accountability fight as DHS's airport face-recognition program, which the ACLU had already sued to expose.
Third-order effects
- If the pattern holds, federal security agencies increasingly source perception AI from a small set of commercial labs, making vendor selection and audit access — not in-house capability — the real lever over how screening systems behave.
- Screening algorithms trained on scanner imagery raise the same consent and bias questions that forced DHS to retreat on citizen face scans, pushing public-safety AI toward formal governance rules rather than case-by-case controversy.
The trend: DHS is institutionalizing its reliance on commercial AI labs — moving from one-off prize contests like this Google partnership to standing task forces and multi-lab pilot programs.