Congress is getting ready to do what it does best: pass a well-intentioned bill that could end up doing real harm.
Buried inside the defense authorization package is the GAIN AI Act, a proposal being sold as a way to strengthen America’s lead in artificial intelligence and prevent advanced US chips from fueling Chinese progress.
The pitch sounds tough: force US chipmakers to prioritize domestic buyers and tighten control of high-end hardware sold abroad.
But in practice, this bill is a self-inflicted wound. It won’t make the United States safer, and it won’t slow China down. Instead, it risks undermining the one arena where hesitation is genuinely dangerous: American leadership in global technology.
Solution in Search of a Problem
The case for the GAIN AI Act rests on the assumption that US chipmakers are neglecting American customers to chase overseas sales. But that simply isn’t true.
NVIDIA, AMD, and Intel have all been explicit: domestic demand comes first. Foreign sales don’t deprive American buyers — they fund the massive R&D pipelines that keep these firms ahead.
There is no evidence of a real domestic shortage that this bill would solve. What it actually creates is a new layer of export restrictions under the banner of “security,” even when the exports in question pose no genuine national-security risk.
This reflects a familiar reflex in Washington: treat every new technology like a weapons system that must be tightly controlled. But innovation doesn’t flourish under a permission-based mindset — and this bill pushes exactly in that direction.
Closing Markets Doesn’t Protect Us — It Hands Them to China
Every time Congress tells an American tech company “you can’t sell that,” Beijing celebrates. Markets don’t remain empty; Chinese firms backed by enormous state investment rush in.
We’ve already seen this dynamic. After the United States restricted high-end chip exports, Huawei didn’t fold. It built its own.
Within a few years, it shifted from a telecom equipment vendor to a credible chip contender, shrinking what was once an insurmountable technological gap.
The same thing happened in AI. When American firms restricted access to their most advanced systems, China simply built its own.
Today, DeepSeek, Alibaba, and Baidu models dominate AI markets across Africa, Southeast Asia, and parts of the Middle East. In places like Nigeria, Kenya, and Vietnam, Chinese systems aren’t alternatives — they’re the default.
This is what happens when the US walls off its own technology: we don’t deny the world advanced tools; we hand the market to Beijing.
Dependence Is Deterrence
Here’s the strategic truth policymakers should keep front of mind: the more the world relies on American technology, the safer the United States is.
If global AI, cloud, and semiconductor ecosystems run on US hardware and American technical standards, we retain visibility, leverage, and influence.
We shape norms, protocols, and the rules of the game. And we understand how adversaries interact with critical technologies because they’re using platforms we built.
That is far stronger than a world in which countries migrate to Chinese systems because Congress pushed them there.
Leading voices in technology policy have warned that aggressive export bans accelerate Chinese self-sufficiency.
Former Google CEO Eric Schmidt has said it plainly. Representative Ro Khanna has gone further: keep the most sensitive defense capabilities out of adversaries’ hands, but keep them reliant on American commercial platforms. That’s long-term leverage.
Ignored Lesson
Protectionism often feels patriotic. It produces tough-sounding headlines. But its track record is remarkably consistent: it weakens the US industrial base instead of strengthening it.
In the 1980s, 100 percent tariffs on Japanese electronics were supposed to save US chipmakers. Instead, they raised prices, snarled supply chains, and allowed South Korea and Taiwan to surge ahead.
More recently, the steel tariffs imposed under the Trump administration were meant to protect American producers. Instead, they increased costs for US manufacturers, triggered retaliation that hit farmers, and delivered only limited gains for domestic mills.
The pattern repeats: restrictions meant to shield American technology end up damaging American competitiveness. The GAIN AI Act fits squarely into that history.
We Know What Doesn’t Work
The 2025 “AI Diffusion Rule” should have been a lesson.
It effectively banned US AI sales to more than 100 countries, including partners with minimal security risk. The result was immediate: confusion among allies, diplomatic frustration, and a rush toward Chinese alternatives.
The next administration wisely reversed course and replaced blunt bans with targeted agreements.
Allowing companies like NVIDIA and AMD to sell chips to China under revenue-sharing arrangements preserved American access to a critical market while giving the US visibility into downstream use. That’s what confident, strategic policy looks like.
GAIN AI would reverse that progress. It signals that America doesn’t trust its own companies and is willing to shrink its technological footprint out of fear rather than strategy.
What Congress Should Do
Congress should reject the GAIN AI Act and pursue an approach rooted in leverage, not limitation:
- Protect genuinely sensitive systems — advanced military AI, cutting-edge lithography tools, and true dual-use technologies.
- Keep commercial export channels open when national-security risks are low.
- Ensure US models, chips, and software remain globally embedded — that’s influence.
- Use targeted, country-specific agreements, not blanket bans, to maintain oversight.
That’s how the United States has led for decades: by being indispensable, not isolated.
The Right Move
Safety doesn’t come from blacklists and bans. It comes from competing — aggressively and at scale.
When global infrastructure runs on US chips, US software, and US standards, the world operates on US terms.
That’s how you win a technological competition: not by pulling back, but by staying on the field and dominating it.
Lt Col (Ret.) Patrick McSpadden served 21 years in the US Air Force as an intelligence officer, with assignments supporting a wide range of aircraft and missions including the F-15, F-16, F-22, A-10, B-1, and Distributed Ground Station.
The views and opinions expressed here are those of the author and do not necessarily reflect the editorial position of Military AI.
