Washington has stopped short of adding DeepSeek to its trade blacklist, even though U.S. officials and rivals have accused the Chinese AI startup of helping Beijing’s security apparatus and skirting chip restrictions. The move keeps more than 100 other firms off the list too, underscoring how the Trump administration is trying to avoid turning a tech fight into a broader break with China.
That restraint is not the same as trust. DeepSeek’s low-cost models jolted the industry in January 2025, and since then the company has faced a growing pile of allegations from U.S. officials and competitors alike. The awkward part for Washington is obvious: the more aggressively it uses trade tools, the more it risks escalating a rivalry it is still trying to manage.
Why DeepSeek stayed off the U.S. blacklist
According to Reuters sources, U.S. officials are holding back on punitive moves because they do not want to inflame tensions with Beijing. DeepSeek was one of the clearest test cases, but it was not alone. CXMT, the Chinese memory-chip maker, was also spared, alongside dozens of other entities flagged as national-security risks.
That fits a familiar pattern in U.S.-China tech policy: accuse first, sanction later, and sometimes not at all if the diplomatic bill looks too high. Export controls on advanced chips have already become a central pressure point, so adding more names to the blacklist would have been a neat way to signal toughness. It also would have made the relationship messier, faster.
The accusations around DeepSeek keep piling up
DeepSeek’s problem is not a lack of enemies. A senior U.S. State Department official told Reuters last year that the company supported Chinese military and intelligence operations, and said it had tried to use shell companies in Southeast Asia to gain access to advanced U.S. chips illegally. Anthropic later said it had uncovered a campaign involving DeepSeek and two other Chinese AI labs to extract capabilities from Claude AI, while OpenAI warned lawmakers that DeepSeek was also targeting its models.
- DeepSeek’s low-cost model made a splash in January 2025.
- Anthropic says DeepSeek and two other Chinese labs tried to mine Claude AI for model improvements.
- OpenAI has told lawmakers DeepSeek is also targeting its systems.
DeepSeek’s V4-Pro price cut adds to the pressure
DeepSeek has not just been making geopolitical trouble; it has been pressuring the business side of AI as well. The company recently cut the price of its flagship V4-Pro model by 75%, a move that looks less like generosity and more like a direct challenge to rivals trying to protect margins in a market that already burns cash for fun.
It also topped a June 2026 ranking of the fastest-growing software suppliers in the enterprise AI services category, which helps explain why both regulators and competitors are paying attention. If U.S. authorities keep leaving DeepSeek off the blacklist, the company gets breathing room. If they change course later, it will probably be because the political cost of restraint finally looks higher than the cost of escalation.