DeepSeek, the latest meteoric mobile AI chatbot app to take the world by storm, has been in the technology hype and government scrutiny headlines lately. DeepSeek released its initial batch of models to the public in November 2023. As a result, it quickly shot to number one in Apple App Store and Google Play charts. The app has come under fire for its connection to state funding and possible bans from devices owned by the U.S. federal government.
Simply associating an entity with these inflammatory and misleading terms — as OpenAI did with DeepSeek — is enough to scare away potential collaborators or home institutions. As worries grow, the U.S. government is weighing a ban on DeepSeek’s models on federal devices, per the Financial Times. New York state officials are on the same page.
Rapid Rise and Controversy
DeepSeek’s rise has been stunning, with the app hitting incredibly high download and usage numbers within days of its launch. The company’s first models —DeepSeek Coder, DeepSeek LLM, and DeepSeek Chat—were presented at the end of 2023. Predictably, the app’s popularity led to a public debate over the ethics of utilizing AI produced with state funding.
Though it was an initial success on the ground, the app’s reputation has enjoyed just the opposite trajectory. Descriptions of DeepSeek’s impact on the industry have run from “upending AI” to being called “over-hyped.” David Carr, editor at Similarweb, highlighted a recent trend, noting that “for March, DeepSeek is in second place, despite seeing traffic drop 25% from where it was in February, based on daily visits.” This dramatic downturn makes it unclear how sustainable this reversal will be given the intense global competition in AI.
Yesterday, the U.S. Commerce Department informed its staff that they should cease using DeepSeek on any government devices. This decision comes on the heels of the state of New York taking steps to ban the tool too. These moves signal a growing concern about DeepSeek’s business practices and ties.
The Next Generation of Models
In late Spring 2024, DeepSeek launched its next-generation model family, DeepSeek-V2, designed to improve performance and usability even further. In line with its new approach to innovation, DeepSeek released an experimental model, V3.2-exp in September 2024. This new model aims to provide a solution with orders of magnitude lower inference costs and high performance expectations.
The company claims that its R1 model is equivalent to OpenAI’s o1 model on key benchmarks. That’s a really encouraging sign that DeepSeek isn’t just chasing the pack, it’s a serious AI contender. With the introduction of DeepSeek-V3 in December 2024, the brand’s marketing stronghold took root. This release really turbocharged its rising fame.
Developers on community platforms such as Hugging Face have flocked to DeepSeek’s tools with an impressive enthusiasm. They have produced over 500 derivative models of R1, leading to an astounding 2.5 million downloads. This high degree of interest points to the fact that the community acknowledges the value and promise of DeepSeek’s technology.
Backing and Technical Team
DeepSeek is the first startup guided by High-Flyer Capital Management. This Chinese quantitative hedge fund is famous for its cutting-edge usage of AI into trading strategies. This support begs further inquiries into the impact of a foreign body’s control over American technology, innovation, and data security.
The technical team in charge of DeepSeek is young and energetic. These new voices and new innovators benefit the growing AI space by providing fresh perspectives and innovative ideas. This youthfulness may contribute to perceptions of naivety regarding regulatory concerns and ethical considerations in deploying AI technologies.
Debates have taken place over how foreign-backed AI companies in operation within the U.S. can endanger data security. State government officials are making sure concerns over a loss of state control come first in these discussions.