Austin, Texas
If Congress wants the United States to fall behind in the global AI race, Oracle’s top policy executive says there is a clear way to do so: pass the Remote Access Security Act.
This is not just speculation. On June 3, 2026, Ken Glueck, Executive Vice President at Oracle AI Infrastructure, issued a strong warning to Washington through Oracle’s official cloud infrastructure channels. He made it clear: the Remote Access Security Act, which is still waiting for Senate approval, is written so broadly that it would almost certainly cause the US to lose what he calls a once-in-a-generation battle for technology leadership. According to Glueck, the stakes are as high as the decision after World War II to make the US dollar the center of global finance.
A company does not make a statement like this lightly.
The Legislation That Could Lock American AI Out of Its Own Future
The Remote Access Security Act (H.R. 2683, S. 3519) passed the House on January 12, 2026, with a strong 369-22 vote. The main goal of the bill is clear: to close a loophole that lets foreign adversaries train AI models on American chips by renting cloud access from overseas data centers. Chinese companies have already taken advantage of this. INF Tech reportedly rented 2,300 Blackwell GPUs through a data center in Indonesia. Tencent signed contracts worth $1.2 billion for 15,000 Blackwell processors through a provider in Japan.
These are genuine security risks. The bill’s sponsors, Senators Dave McCormick (R-PA) and Ron Wyden (D-OR), argued urgently that the Bureau of Industry and Security lacked the legal authority to require a license for this kind of access under current export laws. Fixing this loophole had support from both parties and led to a unanimous 51-0 committee vote in the House.
But Oracle’s executive warning on the Remote Access Security Act cuts through the political consensus with a structural argument that neither party has properly addressed: the legislation is not a scalpel. It is a sledgehammer.
The Overcorrection That Could Hand China the Win
Glueck points out that the bill, as it stands, would block US AI compute providers from 30 to 45 percent of the global market before any contracts are even signed. Instead of targeting specific GPUs, model weights, or large language models, it restricts all cloud services, no matter how harmless the technology might be. It bans all Chinese customers, even if there is no national security risk. Most surprisingly, it also imposes these restrictions on any customer who simply hires a Chinese national.
Glueck offers a clear analogy: imagine telling Apple, Tesla, or Boeing that they cannot do any business with China at all. That is exactly what RASA does to American cloud hyperscalers.
The legislation also leads to confusing results. For example, a Chinese-owned company based in Singapore or the UAE would not be allowed to use US AI compute services, but the same company operating from Shenzhen or Ashburn, Virginia, would be allowed. The bill even affects US allies like the UK and Germany, restricting hyperscale computing for customers with very limited ties to China.
The 5G Warning That Washington Has Already Forgotten
Oracle AI Infrastructure’s concern is based on history, not just theory. Glueck points to Huawei’s dominance in global 5G telecommunications as a key example, and it is worth considering that case closely.
Huawei did not win the 5G race because its technology was better than Western options. Instead, it reached scale first by using government support and low prices to become deeply embedded in global telecom networks, making replacement too expensive. Once a country built its communications backbone with Huawei equipment, switching to Ericsson or Nokia was not just a policy choice—it meant spending billions to overhaul infrastructure. Today, policymakers around the world admit that “ripping and replacing” Huawei is more hope than a real plan.
US AI computing infrastructure is facing a similar turning point. Right now, businesses are choosing their platforms, and governments are picking up long-term AI partners. The global AI market is moving forward, regardless of Washington’s policy debates. If American companies are slower, harder to work with, or blocked from entering big parts of the global market, Chinese competitors like Huawei, DeepSeek, and others will step in. The money they earn will help them build more data centers, fund research, and attract talent, giving them an even bigger advantage in the years ahead.
Why the “Lock It Down” Instinct Gets the Economics Wrong
The Remote Access Security Act is based on the idea that American AI technology is much more advanced than Chinese alternatives, so limiting access would seriously weaken competitors. This made sense for advanced semiconductor manufacturing in 2022, but it is much less clear for cloud AI services in 2026.
As Glueck pointed out, even in areas where the US leads, for example, GPUs and advanced models, the advantage is measured in months rather than years. The basic server hardware, networking, and memory needed for AI have been widely available for a long time. DeepSeek’s rise showed that skilled competitors can get strong results even with limited computing power. In Glueck’s view, “good enough” AI is already enough to secure a lasting place in global markets.
David Sacks, the former White House AI Czar, suggested a different way to measure success: if American technology makes up 80 percent of global computing power in five years, that would be a win. The Oracle executive’s warning on the Remote Access Security Act argues that this law would make reaching that goal impossible. You cannot win 80 percent of a market if you have already blocked yourself from it.
What a Smarter Policy Would Look Like
Oracle does not believe that export controls are always a bad idea. Glueck made it clear that targeted restrictions, focused on enemy militaries, intelligence agencies, sanctioned groups, and diversion networks, are still important tools. The issue is not with the idea of control, but with how precisely it is carried out.
A modern policy for the AI era would distinguish between a Chinese military procurement network and a Singapore-based logistics company with Chinese shareholders. It would use end-user and end-use controls instead of broad bans based on nationality. It would also recognize that Oracle AI Infrastructure and other American companies compete by scale, not scarcity, and that every bit of global market share they lose goes straight to competitors who do not care about American technical standards or governance.
The Internet is a useful example. The United States never fully owned the Internet, but it shaped global technology for 30 years by ensuring that American platforms, protocols, and products became dominant first. As it stands, the Remote Access Security Act would prevent US AI compute providers from repeating this success in the infrastructure that will support future AI.
The Senate Decision That Will Define a Decade
The bill is now waiting for action in the Senate Banking, Housing, and Urban Affairs Committee. Congress’s decision in the next few months will not just settle a disagreement between cloud providers and national security advocates. It will decide whether American AI infrastructure can grow enough to set the technical standards, developer communities, and rules that guide how artificial intelligence works around the world.
Oracle AI Infrastructure‘s warning is a reminder that, within the technology competition, well-intentioned overreach can be just as damaging as inaction. Locking out foreign developers from American-controlled data hubs does not eliminate demand for AI compute. It redirects it — toward rival infrastructure nodes built outside American jurisdiction, operating on non-American standards, and generating the revenue that funds the next generation of American competition.
The Oracle executive’s warning about the Remote Access Security Act amounts to this: security and scale are not opposites, but the current bill treats them as if they were. The Senate has an opportunity to correct that. Whether it takes it will say a great deal about whether the United States intends to lead the AI era or merely participate in it.
Source: How we win.
