U.S. lawmakers today introduced a bill that would prohibit federal employees from using DeepSeek’s app on government devices.

The proposed legislation is known as the No DeepSeek on Government Devices Act. According to Ars Technica, it would ban DeepSeek within 60 days of going into effect. The bill was written by U.S. representatives Josh Gottheimer and Darin LaHood following revelations that DeepSeek’s app is linked to a Chinese state-owned internet provider.

DeepSeek is the Chinese artificial intelligence lab behind R1, an advanced large language model that debuted last month. The LLM has drawn significant interest because it can outperform OpenAI’s o1 reasoning model across several tasks. Moreover, DeepSeek claims that it trained the model using only $5.6 million worth of graphics card hours.

The AI developer has released a free app that provides access to DeepSeek-V3, a predecessor of R1. Both models have a mixture-of-experts design with 671 billion parameters. However, R1 was developed through an enhanced training workflow that makes it better at tasks such as solving math problems.

Last month, DeepSeek overtook ChatGPT as the most downloaded free mobile app in the U.S. The company also offers a web version of its app. The latter service was recently the focus of a high-profile cybersecurity investigation, which is what led to today’s legislative proposal. 

Earlier this week, Canadian cybersecurity company Feroot Security revealed that the web version of DeepSeek’s app contains a “heavily obfuscated” script. That means the script was written in a way that makes it difficult for researchers to understand how it works. Feroot determined that the code connects to the infrastructure of China Telecom, a Chinese state-owned telecommunications company that is banned from operating in the U.S. 

The script in question is embedded in DeepSeek’s login page. The code is believed to be connected to authentication and identity management systems operated by China Telecom. After users sign in, the code collects detailed technical data about their devices through a process known as fingerprinting.

During its analysis, Feroot Security did not detect attempts by the app to send user data to China Telecom systems. However, some cybersecurity researchers believe it’s likely that such data transmission is taking place.

DeepSeek was already facing scrutiny before today’s legislative proposal. In recent weeks, the U.S. Navy and NASA have banned employees from using the app. DeepSeek is likewise banned on Texas state government devices.

The app has also raised cybersecurity concerns elsewhere. South Korea and and the Netherlands have banned DeepSeek on government devices. Last week, Italy’s data privacy watchdog went a step further and made the app inaccessible for consumers. 

Image: Unsplash