OpenAI slams DeepSeek, warning the US government that China is catching up fast

• In a letter to the US government, OpenAI slammed its fast-rising Chinese rival, DeepSeek.
• "While America maintains a lead on AI today, DeepSeek shows that our lead is not wide and is narrowing," OpenAI wrote.
• The letter also outlined policy recommendations to secure America's lead in AI.

OpenAI has launched an attack on its fast-rising Chinese rival, DeepSeek.

In a 15-page letter to the US government on Thursday, OpenAI called DeepSeek's latest model, R1, a "noteworthy" development — one that signals China's growing AI ambitions and underscores the tightening competition between the rival countries.

"While America maintains a lead on AI today, DeepSeek shows that our lead is not wide and is narrowing," Chris Lehane, OpenAI's vice president of global affairs, wrote to the Office of Science and Technology Policy.

Founded in 2023 by Chinese entrepreneur Liang Wenfeng, DeepSeek shook up the AI industry and the US stock market with its low-cost reasoning model, R1, unveiled in January. The company said its R1 model rivals top competitors, like ChatGPT's o1, but at a fraction of the cost.

The company behind ChatGPT didn't hold back its criticism of the Chinese model. Lehane warned that using DeepSeek in critical infrastructure and other high-risk applications poses a "significant risk" because DeepSeek could be pressured by the Chinese government to manipulate its models.

The US Navy has warned service members against using DeepSeek, and Taiwan banned it in government agencies in February over security concerns.

"Because DeepSeek is simultaneously state-subsidized, state-controlled, and freely available, the cost to its users is their privacy and security," Lehane wrote.

Lehane also said DeepSeek's AI is "more willing" to generate responses for illicit and harmful activities, including identity fraud and intellectual property theft.

"The Chinese Communist Party views violations of American intellectual property rights as a feature, not a flaw," he wrote.

Lehane warned that China would use AI as a geopolitical tool by offering DeepSeek to countries needing AI tools and infrastructure funding — a move aligned with its existing Belt and Road initiative. For over a decade, China has used this program to spend more than $1 trillion on infrastructure programs globally.

OpenAI's CEO Sam Altman called DeepSeek's R1 an "invigorating" competitor in January, saying it would push his company to release "better models" faster.

Securing America's lead in AI

OpenAI's letter also outlined policy recommendations to secure America's lead in AI.

The letter was written in response to the White House Office of Science and Technology Policy's request for public input in February.

In January, President Donald Trump signed an executive order calling for the US to "sustain and enhance America's global AI dominance" while revoking the Biden administration's order demanding greater transparency from AI companies.

His administration is now tasked with developing an AI Action Plan by July.

Lehane on Thursday proposed a series of "freedom-focused" policies, including a regulatory strategy that would relieve American AI developers from complying with "overly burdensome state laws."

The company also said the US government could make it easier for AI companies to train on copyrighted material, arguing that strict copyright laws could slow innovation and restrict access to training data.

OpenAI and its peers have been criticized for using copyrighted content to train models. Authors and news outlets, including The New York Times, have sued OpenAI, saying the startup is breaking copyright law.

Lehane proposed a copyright strategy that would protect creators' rights while safeguarding America's AI leadership and national security.

"The federal government can both secure Americans' freedom to learn from AI, and avoid forfeiting our AI lead to the PRC by preserving American AI models' ability to learn from copyrighted material," he said.