)
tanarch – stock.adobe.com
US lawmakers have introduced a bill to ban the DeepSeek AI on government-owned devices. They cited national security concerns because of its alleged links with the Chinese state.
Alex Scroxtonis the Security Editor.
US legislators in Washington DC this week voted to ban the use of DeepSeekis the breakthrough Chinese generative artificial Intelligence (GenAI) tool which swept the world and wiped out billions of dollars from the value of US technology companies at the end January.
No DeepSeek on Government Devices Act, a bipartisan legislation, was introduced by Democratic Congressman Josh Gottheimer, and his Republican counterpart Darin LaHood. They represent districts in New Jersey and Illinois, respectively.
This legislation will attempt to ban the download and use of DeepSeek AI software on government-owned devices. Taiwan and the Australian government have taken similar steps. Other countries are also considering taking action. The Italian data protection authority announced that DeepSeek would be limited in its processing of data from Italian users.
In America, several bodies, including the State of Texas and the US Navy, have already taken steps to ban the app.
The US’s efforts to restrict DeepSeek’s use are similar to the controversial TikTok ban – currently on pause for 75 days after an executive order signed President Trump – and reflect the Western bloc’s long-held concern over the Chinese government’s ability to co-opt user data from technology organisations at will.
LaHood said that the national security threat DeepSeek, a CCP [Chinese Communist Party]-affiliated firm, poses to the United States was alarming. “DeepSeek’s generative AI program collects data from US users and stores it for use by the CCP in an unidentified manner.
We will not allow a CCP firm to access sensitive government or personal information. This bipartisan, common-sense piece of legislation will prohibit the app from federal employees’ phones and close backdoor operations that the company is trying to exploit in order to gain access. It is crucial that Congress safeguard Americans’ data and continue to ensure American Leadership in AI.”
Gottheimer continued: “The Chinese Communist Party made it abundantly obvious that it would exploit any tool available to it to undermine our national safety, spread harmful disinformation and gather data on Americans. We have disturbing evidence that the Chinese Communist Party is using DeepSeek in order to steal sensitive data from US citizens.
This is a national security fire of five alarms. We must find out the truth about DeepSeek’s malicious activities. We cannot risk the CCP infiltrating our government officials’ devices and compromising our national security. We’ve seen China playbook before, with TikTok. And we cannot allow it again.”
For many years, there has been a high level of concern about the use of nonallied technology by government and military agencies. Mel Morris, CEO at AI research engine Corpora.ai .
One could argue that it is a prudent measure that ensures that devices cannot by compromised by a possible adversary. In the event of conflict, there is no rule, so any assurance or level of confidence that may exist will likely be thrown out the window. I suppose that the old saying ‘all is fair …’.’ applies. Many would argue that running a service with DeepSeek using an air-gapped system could be made safe and be considered as such.”
Ilia Kolochenko (19659024): Thinking beyond China (). ImmuniWebCEO, and BCS Fellow, said that while the risks arising from the use DeepSeek are reasonable and justified, politicians could miss the forest for trees and should expand their thinking beyond China.
He said that many other GenAI vendors, as well as global SaaS platform providers who are rapidly integrating GenAI capabilities, have similar or worse problems.
While DeepSeek’s risk should not be discounted, we must also remember the fundamental risks and issues of all other GenAI providers. Many of them have unwarrantedly scraped proprietary and copyrighted internet content to train their powerful LLMs, without ever asking permission from content creators or owners. They now vigorously deny any wrongdoing on a variety of untenable pretexts.
Kolochenko said that the DeepSeek incident should not be used as an excuse to ignore serious AI-related risks and violations by other GenAI vendors.
By Aaron so.
DeepSeek R1 generates more dangerous content than other large-language models
by: Alex Scroxton.
