Image credit: Grok/X

Washington is scrambling to get a handle on DeepSeek, a Chinese AI firm that stunned the tech world this month by rolling out a model seemingly on par with industry leaders like OpenAI—at a fraction of the cost.

The kicker? They gave away the code for free.

Now, top White House officials are sounding the alarm over whether DeepSeek has quietly leveraged a controversial method called "distillation," a technique that allows one AI model to learn from another—potentially piggybacking off American-made innovations.

In Silicon Valley, where AI breakthroughs are fiercely competitive, sources say stopping distillation in its tracks is easier said than done.

The process works by having a powerful, established AI model evaluate and refine the output of a smaller, newer model, effectively passing down its knowledge.

That means a company like DeepSeek could reap the benefits of the billions poured into AI development by U.S. firms—without having to foot the bill.

While distillation is a widely used technique, it also happens to violate the terms of service of leading U.S. AI companies, including OpenAI.

READ MORE: China claims DeekSeek under cyberattacks

Nvidia plummets 14%

DeepSeek overtakes ChatGPT in Apple Store

Meet Liang Wenfeng, the visionary behind DeepSeek

The ChatGPT maker acknowledged it's aware of groups in China actively working to replicate its models through distillation and is investigating whether DeepSeek improperly tapped into its technology.

Naveen Rao, an AI expert and vice president at San Francisco-based Databricks, noted that borrowing ideas—even from competitors—is nothing new.

"This happens in every industry," he said, comparing it to car manufacturers dissecting rival engines. "We all try to be good citizens, but we're all competing at the same time."

Howard Lutnick, former President Donald Trump's pick for Secretary of Commerce, didn’t mince words when addressing DeepSeek at his Senate confirmation hearing:

"I do not believe that DeepSeek was done all above board. That's nonsense."

Meanwhile, David Sacks, the White House’s AI and crypto czar, raised similar concerns on Fox News, warning that distillation could give China an unfair advantage.

OpenAI, for its part, said it will work with the U.S. government to protect American AI technology, though it didn’t specify how it plans to enforce safeguards.

The situation echoes past efforts to curb China's access to key U.S. tech, particularly in semiconductors, where Washington has restricted exports of high-end chips and manufacturing tools. But when it comes to AI, blocking distillation may be like trying to catch smoke with bare hands.

Tech insiders say stopping distillation outright is nearly impossible, especially when AI models like Meta’s Llama and French startup Mistral’s are freely available.

"It's impossible to stop model distillation when you have open-source models," said Umesh Padval of Thomvest Ventures. "They are available to everybody. They can also find OpenAI's model somewhere through customers."

Meta does require companies to disclose if they use Llama for distillation, but whether DeepSeek followed that rule is unclear. The company admitted to using Llama in some of its recent models but didn’t specify if it had done so earlier in the process.

A source at a major AI lab suggested the only real way to prevent firms like DeepSeek from siphoning U.S. AI knowledge would be strict "know-your-customer" regulations—similar to how banks verify clients. But while the Biden administration floated such measures, it’s unclear if Trump, should he return to the White House, would push them forward.

One company taking matters into its own hands is Groq, an AI computing firm that has outright blocked all Chinese IP addresses from its cloud services. But CEO Jonathan Ross acknowledged that’s hardly a foolproof solution. "People can find ways around it," he admitted.

"It's a cat and mouse game … I don't know what the solution is. If anyone comes up with it, let us know, and we'll implement it."

For now, DeepSeek remains silent on the controversy. But with Washington and Silicon Valley on high alert, this battle over AI supremacy is just getting started.

‍