No expectant moms at the border: Trump's birthright Plan B - Axios

When Axios broke the story of "No expectant moms at the border: Trump's birthright Plan B," the immediate reaction centered on constitutional law and political strategy. But beneath the headlines about the 14th Amendment and executive orders lies a far more interesting technical story-one about data pipelines, identity verification systems. And the engineering challenges of rearchitecting citizenship documentation on the fly. The executive order to end birthright citizenship wasn't just a legal gamble; it was a data infrastructure problem waiting to explode.

As a software engineer who has built identity management systems for government health programs, I can tell you that the technical debt required to add a "Plan B" for birthright citizenship is staggering. The current system-which automatically grants citizenship to anyone born on U. S soil, regardless of parental immigration status-is deeply embedded in federal and state databases, from Social Security Number issuance to passport applications. Changing that would require rewriting the logic of dozens of interconnected systems, each with its own data model, legacy code. And failure modes.

This article isn't about the political merits of the policy. It's about what happens when a government tries to implement a sweeping change to a fundamental legal status using systems designed for a completely different set of rules. We'll explore the engineering behind identity verification at the border, the algorithmic challenges of determining "jurisdiction" in real time, and why the tech industry's silence on this issue is deafening.

The Data Pipeline That Defines Citizenship

When a child is born in the United States, a cascade of automated events triggers. The hospital transmits a birth certificate to the state vital records office. Which sends data to the Social Security Administration (SSA) and the Department of State for passport eligibility. Each step relies on a binary assumption: born in the U, and s equals citizenThere is no flag for "parental status" in most of these systems because the 14th Amendment's Citizenship Clause has been interpreted as categorical for over a century.

Under Trump's proposed Plan B-which Axios reported as a strategy to bar pregnant women from crossing the border-the technical challenge flips from "record citizenship" to "determine parentage and jurisdiction before recording citizenship. " That requires a real-time decision engine that can evaluate the legal status of each parent, their location at birth. And the constitutional nuance of "subject to the jurisdiction thereof. " No existing system in the federal government has that capability without massive retooling.

In production environments, we call this a "breaking change" to an API contract. The downstream consumers of birth records-hospitals, state DMVs, the SSA-all expect a consistent, immutable identifier (the SSN) linked to a citizen. If you start issuing conditional or deferred citizenship, you break every interface that assumes the record is final. The ripple effect would be felt for decades.

Machine Learning Models for Fraud Detection at the Border

Border enforcement already uses machine learning to detect fraudulent documents and identify security risks. Customs and Border Protection (CBP) runs predictive models on traveler manifests, visa applications, and biometric data. Adding a "birth tourism" detection layer is technically feasible but fraught with false positives. Imagine an algorithm that flags every pregnant woman arriving from a country with high visa overstay rates-that's the logical extension of "no expectant moms at the border. "

A well-known 2019 CBP pilot used machine learning for border security risk assessment that analyzed 20+ variables including travel patterns, financial data. And social media activity. Recalibrating such a model to predict "future birth citizenship" would require labeled training data (which doesn't exist because the policy hasn't been implemented) and would almost certainly embed racial and socioeconomic biases. The Algorithmic Justice Initiative at the ACLU has documented how similar models have disproportionately flagged minority travelers.

Furthermore, the latency requirements are unrealistic. Border agents need decisions in seconds, not minutes. Adding a citizenship prediction layer to an already strained infrastructure-CBP processes over a million people daily at airports alone-would degrade system performance. We saw this with the 2017 travel ban rollout. Where rushed implementation caused chaos at ports of entry.

Real-Time Analytics for Policy Implementation: A Case Study in Failure

Government IT projects are notorious for overruns. The GAO's high-risk list includes multiple DHS IT systems that are decades old, running on mainframes. And lack documentation. Adding a new adjudication layer for birthright citizenship would be like trying to install Kubernetes on a Commodore 64. The proposed Plan B likely involves a centralized database of "birth tourism" flags, but centralized identity databases have a terrible track record. The FBI's Next Generation Identification system took over a decade and billions of dollars.

From a data engineering perspective, the hardest part isn't the algorithm-it's the data consistency. How do you determine if a pregnancy began before the mother entered the U. S. Medical records from foreign hospitals are rarely digitized or standardized. How do you verify the father's immigration status if he isn't present? The system would need to query multiple federal databases-USCIS case files, DHS entry/exit records. And possibly foreign government databases-with varying degrees of reliability. This is a classic "distributed transaction" problem, and we all know how CAP theorem applies: you can't have consistency, availability. And partition tolerance simultaneously. The government would likely choose availability (process quickly) over consistency (accurate decisions), leading to wrongful denials of citizenship.

In my own work with federal identity systems, we learned that any rule change with a false positive rate above 0. 1% creates a crisis of legitimacy. For a policy affecting families and newborns, even a 0. 1% error rate would impact thousands of children annually. The engineering discipline required-robust error handling, audit trails. And appeals workflows-is absent in most legacy government systems.

Blockchain for Immutable Citizenship Records, but

Some technologists have proposed using blockchain as a decentralized ledger for citizenship data? The idea is tempting: use a tamper-proof distributed system to record births and parental status, making fraud harder and data more transparent. But this misunderstands both the policy problem and the technology. Citizenship isn't a token you can mint-it's a legal status that requires human judgment and due process. Smart contracts can't interpret the "subject to the jurisdiction" clause.

Moreover, blockchain doesn't solve the identity verification problem. You still need an oracle to input the initial data-and that oracle (a hospital, a border agent, a clerk) is the weakest link. If the oracle is corrupt or error-prone, the ledger is garbage-in-garbage-out. The GAO has studied blockchain for public records and found it impractical for complex legal statuses like citizenship due to governance and legal recognition challenges.

Instead of blockchain, the government would be better served by improving existing data integration tools. APIs that allow seamless querying of USCIS and DHS databases, with proper consent and privacy safeguards, could handle most of the decision logic. But that requires months of engineering sprints, cross-agency agreements, and congressional funding-all of which are in short supply when policies are announced via tweet.

The Ethical Engineering of Automated Decision-Making at Scale

As engineers, we have a responsibility to consider the second-order effects of systems we build. The "no expectant moms at the border" frame inherently discriminates against all pregnant women, regardless of their intent. An algorithm that denies entry based on pregnancy alone will inevitably be gamed (e g., non-pregnant women carrying surrogates, false negatives from early-term detection). More importantly, it sets a precedent for using predictive models to strip constitutional rights before they're even earned.

The Electronic Frontier Foundation's guidelines on responsible AI emphasize transparency, accountability. And redress. A citizenship determination system would need to provide clear explanations for denials and allow for human review. Today's CBP systems provide neither. The few public reports on CBP's use of AI reveal black-box models with no audit trail. That's unacceptable for a decision that changes a person's entire legal identity.

In the private sector, we would never deploy a system that affects millions of people without months of load testing, A/B testing. And fallback mechanisms. Yet the government regularly rolls out life-changing policies with less rigor than a startup shipping a weekend project. This isn't about being anti-innovation; it's about demanding the same engineering discipline from policymakers that we demand from our own teams.

Lessons from Production Systems for Policy Execution

If Trump's Plan B were a software project, it would be canceled within the first sprint. The requirements are contradictory: deny citizenship to certain births while not violating the 14th Amendment (which the Supreme Court has repeatedly upheld). The stakeholders are hostile (Congress, state governments, advocacy groups). The data is incomplete. And the timeline is arbitraryNo product manager worth their salary would sign off on that scope.

Yet governments continue to launch policies with the launch-and-learn mentality of a startup. The result is what we saw with the Census citizenship question: legal challenges, court injunctions. And a mess that takes years to untangle. The Axios report itself cites internal DHS warnings about the operational feasibility of denying entry to pregnant women. This is the equivalent of a failed integration test: the system cannot handle the load, the data formats don't match. And the error handling is nonexistent.

What policymakers need-and what the tech community should advocate for-is a phased approach. Start with a pilot program at a single port of entry, using manual reviews alongside automated suggestions. Collect data on false positive rates - processing times, and legal challenges, and iterate based on feedbackThis is basic DevOps, but it's rarely applied to legislation. Until Congress mandates a tech impact assessment for any immigration policy change, we will keep repeating these failures.

FAQ: Birthright Citizenship and Tech Implementation

Q1: Could the government deploy a real-time decision engine at every border crossing?
A: Theoretically, yes. But it would require a nationwide network upgrade, standardized data formats. And integration with state birth registries. The cost estimate for a similar project-the DHS's failed ATS system-was over $1 billion. CBP's current network has latency and bandwidth limitations in rural border areas.

Q2: Would machine learning models for citizenship eligibility be biased?
A: Almost certainly. Training data would come from historical border enforcement. Which disproportionately targets minority travelers. Without explicit debiasing techniques, any model would reflect historical discrimination. The ACLU has already filed lawsuits over similar CBP algorithms.

Q3: Can blockchain solve identity verification for birthright citizenship?
A: No. Blockchain provides immutability but not truthfulness. The input data (a mother's immigration status) is subjective and can be falsified. Smart contracts can't interpret the 14th Amendment. The technology is a solution in search of a problem for this use case.

Q4: How long would it take to build a system that determines citizenship at birth?
A: At least 5-10 years, given the need for legal approval, interagency agreements - data standardization. And testing. The Electronic Verification of Employment (E-Verify) system took over a decade to achieve nationwide coverage for a simpler task: confirming work authorization.

Q5: What happens to children born during the transition period if the policy is implemented?
A: This is the hardest engineering challenge-backwards compatibility. Existing records would need to be retroactively re-evaluated, a process that would require massive data migration and likely result in many incorrect reclassifications. The SSA alone would need to reissue millions of SSNs.

Conclusion: Why Engineers Must Speak Up

The "No expectant moms at the border: Trump's birthright Plan B - Axios" story is more than a political flashpoint. It's a textbook case of what happens when policy outpaces infrastructure. As technologists, we have a responsibility to call out technically infeasible proposals before they cause harm. We can build identity systems. But we can't build systems that legislate the Constitution. The engineering community must demand that any policy change involving large-scale data collection and decision-making undergo a rigorous technical review, with public disclosure of the architecture, algorithms. And failure modes.

If you work on identity systems, border technology. Or government contracts, your voice is needed now more than ever. Write your representatives. Submit public comments on DHS rule changes, and share this article with your networkThe algorithms we build today define the citizenship of tomorrow.

What do you think?

Should there be a mandatory "tech impact statement" filed for every immigration policy change, similar to environmental impact statements?

Given the high false-positive rates of current algorithmic systems, do you believe any automated citizenship determination can be fair and equitable?

Would you feel comfortable deploying a machine learning model that could deny constitutional rights based on predictions-and if so, what safeguards would you demand?