As DeepSeek Soars, Many Legal, Ethical, And Social Questions Arise About Its Development

It’s been over 10 days since Chinese artificial intelligence (AI) startup DeepSeek revealed its R1 AI model was at par with OpenAI’s latest models—even better than in some cases—and the headlines continue to be dominated by the new player in the field.

However, the headlines now are raising questions about the AI model, its validity, and whether it poses a threat to individuals and national security.

For one, DeepSeek claimed to have built its almighty family of models using 2,048 lower-powered NVIDIA H800 GPU chips and just $5.6 million.

This claim is being challenged by numerous experts, businesses, and even the United States Department of Commerce, as per an anonymous source.

The U.S. DOC is investigating if DeepSeek sources chips from other countries, circumventing the existing trade restrictions put in place that prevent Chinese companies from acquiring high-powered and sophisticated chips amid the current AI war.

As per the source, DeepSeek may have sourced restricted chips using the organized chip smuggling channel that’s been tracked out of countries such as Malaysia, Singapore, and the United Arab Emirates.

While the U.S. DOC and DeepSeek didn’t provide a comment, an NVIDIA spokesperson said the company is enquiring with its customers, who have business entities in Singapore, adding, “We insist that our partners comply with all applicable laws, and if we receive any information to the contrary, act accordingly.”

Reportedly, DeepSeek also “extensively used” NVIDIA H20s GPU chips, something that U.S. lawmakers are urging President Donald Trump to include in the export curbs.

“We ask that as part of this review, you consider the potential national security benefits of placing an export control on Nvidia’s H20 and chips of similar sophistication,” wrote U.S. Congressional members Republican John Moolenaar and Democrat Raja Krishnamoorthi in a letter addressed to National Security Advisor Michael Waltz.

As for national security, the U.S. House of Representatives' Chief Administrative Officer, Catherine Szpindor, has issued a notice restricting congressional offices from using DeepSeek.

An excerpt from the notice reads, “At this time, DeepSeek is under review by the CAO and is currently unauthorized for official House use.”

The move came soon after it was revealed that U.S. Department of Defense workers tried out DeepSeek over two days, which connected their computers to Chinese servers. This led to The Pentagon blocking DeepSeek on some of its networks.

Furthermore, the US Navy also banned DeepSeek, and other generative artificial intelligence (GenAI) applications, such as ChatGPT, and even warned its people from using such apps, citing “potential security and ethical concerns associated with the model's origin and usage.”

A navy official said, “The internal correspondence was a reminder to colleagues of standing Navy guidance against the use of any publicly accessible, open-source AI programs or systems for official work,” where “DeepSeek was mentioned as the most recent example of how the standing guidance applies.”

It’s not just the U.S. that’s questioning the global GenAI sensation.


Privacy watchdogs in South Korea, France, Italy, and Ireland are seeking answers about DeepSeek’s personal information use and data protection measures, as it looks to understand how its AI system works and what possible risks could it hold for users. Additionally, the question if DeepSeek violates Europe’s General Data Protection Regulation (GDPR) arose.

The hunt to determine how DeepSeek stores and manages the personal data of users may be a justified move, considering security researchers from cloud security firm Wiz found that DeepSeek operates an open-source data management system called ClickHouse.

This publicly accessible database allows anyone to access data such as chat histories, API authentication keys, system logs, and other sensitive material within minutes, without authentication. The database also consisted of more than 1 million log lines.

Wiz noted that the exposure “allowed for full database control and potential privilege escalation within the DeepSeek environment.”

After being notified by Wiz, DeepSeek secured its database.

On the other hand, OpenAI and Microsoft—its primary backer—accused DeepSeek of plagiarizing portions of OpenAI’s code and its proprietary models to train its system through distillation. Distillation is when one model learns from another.

Allegedly, DeepSeek also extracted OpenAI’s data through its API.

OpenAI drew criticism from users from around the world for being bitter at DeepSeek’s success, seeing how OpenAI also built its models on free access of copyrighted data. OpenAI has been sued by various publishers for using their data to train their models without authorization.

Furthermore, DeepSeek has been built using much fewer resources than OpenAI did, especially monetarily.

Interestingly, a group of researchers at the University of California, Berkeley, claim to have recreated DeepSeek’s core technology for much, much less than what DeepSeek did—just $30.

UC Berkeley is also a contributor to the newly launched Oumi, which is the world’s first unconditionally complete AI platform that’s looking to do what DeepSeek did.

The researchers replicated DeepSeek-R1-Zero’s reinforcement learning capabilities using a small language model (SLM) with just 3 billion parameters.

Do you think users need to be more careful about DeepSeek? Do you think DeepSeek should be investigated to ascertain if the company sourced and used restricted AI chips, or is their method of acquiring them fair game?

Let us know in the comments below!