DeepSeek's AI Agent: A New Challenger in the Global AI Race

DeepSeek Is Working on an AI Agent: Will It Be Better Than ChatGPT?

China-based DeepSeek is working on developing a new agentic generative AI model, Bloomberg reports, citing anonymous sources. Agentic AI is the latest wave of AI technology. AI agents are a kind of digital assistant; they can complete tasks without a lot of human oversight. AI agents can do anything from coding to ordering you a pizza, as my colleague Imad Khan recently tested. Details about the specifics of the DeepSeek agent model are still fuzzy. An August update to DeepSeek's V3 model was a "first step toward the agent era," according to an X post by the company. Like any AI tool, a future DeepSeek agent's usefulness will be contingent on its availability and price. DeepSeek did not immediately respond to a request for comment. Don't miss any of our unbiased tech content and lab-based reviews. Add CNET as a preferred Google source. Many AI companies have been investing in agentic AI, releasing ready-made features like ChatGPT Agent and Copilot agents. So it's not surprising to hear that DeepSeek is working on its own version of an AI agent. But it does raise a lot of questions about how that will affect the global AI landscape and ultimately the tools we have access to and use. DeepSeek threw American AI companies for a loop at the beginning of the year when it released its R1 reasoning model. The AI stunned chatbot users with its behind-the-scenes look at how the machine "thinks" and processes requests. DeepSeek lapped American rivals like OpenAI, Microsoft and Meta with the model, showing it could not only make quality AI models at a fraction of the cost of American companies, but that it can make them more open, too. Tech companies' stocks plummeted, with AI chip-maker Nvidia seeing its market cap drop by $600 billion in a single day. Beyond showcasing the technical prowess of Chinese models, it also spurred a lot of national security concerns and fueled the desire to win the global AI race. Concerns about data privacy abound with any AI model, but the specific concern with DeepSeek is that American users' data would end up under the control of the Chinese government. Both the Biden and Trump administrations sought to ensure the US would win the AI race -- that is, have the most advanced AI tech controlled solely by American companies. Chinese models like DeepSeek helped fuel the Trump administration's AI action plan, which proponents claim greenlights AI innovation by removing bureaucratic red tape. Critics worry that it could erode protections for consumers and the environment.

DeepSeek may be about to shake up the AI world again - what we know

DeepSeek, the Chinese AI startup that sent shockwaves throughout Silicon Valley earlier this year with its sudden ascent onto the global tech scene, is reportedly gearing up to launch its most powerful AI system yet. The company aims to release an AI agent to compete with similar models from OpenAI, Google, and other tech giants by the end of this year, according to a Thursday report from Bloomberg, which cited anonymous sources. Also: What is sparsity? DeepSeek AI's secret, revealed by Apple researchers In contrast to traditional chatbots, which require consistent prompting from human users, agents can autonomously execute multi-step tasks with little to no human oversight. Tech developers have been promoting them as the next rung in AI's evolutionary ladder, and as productivity boosters for businesses looking to automate some mundane and time-consuming aspects of everyday work. In January, DeepSeek released R1, an AI model that quickly gained international attention for its ability to clearly describe to users how it was working through its reasoning process to arrive at its responses to queries, like a precocious student showing all their work on a math test. More impressive, however, was the alleged cost required to build it: DeepSeek reportedly only spent around $6 million to build R1 -- a relatively paltry amount compared to the estimated costs of developing ChatGPT, Gemini, and Claude. And unlike those other models, R1 was open source, meaning developers can access its underlying code and build their own applications on top of it for free. Also: I tested DeepSeek's R1 and V3 coding skills - and we're not all doomed (yet) The sudden appearance of R1 seemed to upset the "bigger is better" paradigm, which by then had become deeply entrenched in Silicon Valley and is based on the premise that advanced models require gargantuan quantities of data -- and money -- to build. For the first time, here was a model whose performance seemed to rival leading LLMs, and whose construction required far fewer resources than were being spent by its competitors. It's been called China's Sputnik moment. Not much is known at this point about DeepSeek's reportedly forthcoming agent. But if recent trends across the AI sector are any indication, it's probable that it will be promoted as an automated virtual assistant for professionals and businesses, which has become an overarching marketing theme for most agents currently on the market. Also: The best AI chatbots of 2025: ChatGPT, Copilot, and notable alternatives Regardless of how the model ultimately turns out, Silicon Valley will certainly be keeping a close eye on developments from the Hangzhou-based company. For the time being, the bigger-is-better paradigm is still reigning throughout tech, and companies like OpenAI, Google, Microsoft, and Meta are investing in ever more massive data centers to fuel their swelling AI goals. The appearance of a less data-hungry agent from DeepSeek, which, perhaps like R1, could cost less to build than other popular agents, could send yet another seismic shock throughout the tech sector. Washington is also likely to be warily awaiting further news from DeepSeek; the sudden and tumultuous arrival of R1 earlier this year rattled US policymakers who fear China could gain a competitive edge in the AI race.

DeepSeek Targets AI Agent Release by End of Year to Rival OpenAI

DeepSeek is developing an artificial intelligence model with more advanced AI agent features to compete with US rivals like OpenAI in a newer frontier of the technology, according to people familiar with the matter. The Hangzhou-based startup is building an AI model that's designed to carry out multi-step actions on a person's behalf with minimal direction from the user, said the people, who spoke on condition of anonymity as the information is private. The system is also meant to learn and improve based on its prior actions, the people said.