DEEPSEEK’S debut has sent ripples through Silicon Valley, Wall Street, and Washington not only because of its performance vis-à-vis the current mainstays, Open Source ChatGPT, Google Gemini and Microsoft Co-Pilot but also because it was supposedly developed at a significantly lower cost.
DeepSeek also triggered many security and privacy fears, conjuring a lot of conspiracy theories, especially about why it doesn’t answer questions about Chinese history or personalities.
There are two versions of DeepSeek available. The downloadable app version called DeepSeek-AI Assistant (also called DeepSeek V3) which is readily available online. Then there is an open source version called DeepSeek R1, which has an open-source architecture, which is available on platforms like GitHub.
This accessibility enables developers and researchers to modify and customize the model according to their specific needs. Unlike traditional app-based models that often require ongoing subscriptions, DeepSeek-R1 is free to use, making it an attractive option for individuals and organizations looking to explore AI without financial barriers.
Fitering
Reports that have come out about DeepSeek being “politically motivated” especially about questions about China, on sensitive topics like Tiananmen Square, the West Philippine Sea and Xi Jin Ping.
According to a discussion on the TechSabado, Democracy.Net.PH convenor Carlos Nazareno said that this is because most do not understand the two aspects of DeepSeek. The app-based version is subject to Chinese legal and regulatory compliance, meaning AI platforms must operate within Chinese legal frameworks, which may restrict discussion of sensitive topics in specific regions.
DeepSeek’s product development online literature pointed out that on the app version, the filtering is used to “avoid spreading unverified claims or politicized narratives thus the app refrains from engaging in topics that require subjective interpretation or lack authoritative, publicly verifiable sources.”
Unfiltered
On the DeepSeek R1 open source version however, things are very different. Aside from being very technical, lacking the premium, easily-accessible qualities of the app, it is also free to use but charges user small amounts of money for server usage and space.
DeepSeek-R1 runs under an MIT license allowing for free commercial and academic use. The MIT license is permissive, granting users broad freedoms to utilize, modify, and distribute software, with the sole requirement being the preservation of the original copyright notice and license text. This simplicity and flexibility make it a popular choice for AI developers, fostering a collaborative environment where models, datasets, and tools can be readily shared and built upon.
Allowing direct access from different AI platforms and give programmers and developers direct access to the huge lake of information on the internet. This makes for ‘unfiltering’ of the data and the way data is searched and analyzed. In this case, no filtering is done.
Since it optimized for efficiency, it can run on consumer-grade hardware like laptops. The model’s architecture, quantization, and lightweight design allows for this. Optimization is also the reason why DeepSeek has aligned other open source platforms like Meta’s Llama or Mistral AI.
Daniel Winter, DW Business editor said in an DW podcast that for DeepSeek “you need to throw everything you know about AI out the window.” He cited that apart from developing an AI chatbot for the fraction of the cost, the platform also allowed learning and development from other AI models.”
Artificial general intelligence
Winter also mentioned how tech author Ben Thompson said that because of DeepSeek, “we are close to AGI (Artificial General Intelligence),” which is a new stage in artificial intelligence development.
AGI is a hypothetical form of AI that possesses human-like general intelligence—the ability to learn, reason, adapt, and perform any intellectual task a human can do. Unlike today’s narrow AI like search engines, ChatGPT or autonomous cars, which excel at specific tasks, AGI would be able to think abstractly and understand context across domains.
Moreover, AGI can transfer knowledge from one field to another—an example if to compose music using mathematics or create artwork using chemical equations. This transfer knowledge skill isn’t a creative skill but rather using computational skills to develop abstract executions.
What is most disturbing about AGI, according to AI scientists, is it can learn autonomously without human intervention and possibly later on, may exhibit what can be defined consciousness, even self-awareness, or intentionality.
Wake-up call
US President Trump called DeepSeek’s rise to popularity as the most downloaded app in the US as a “wake-up call” for American technology firms. Trump emphasized the need for U.S. companies to adopt a more competitive mindset, urging them to innovate and reduce costs to keep pace with their Chinese counterparts.
He pointed out that DeepSeek’s ability to deliver high-quality AI services at a fraction of the cost but also touched on broader implications for national security and regulatory policy. While he did not propose specific restrictions on AI technologies, there was a clear acknowledgment of the geopolitical stakes involved.
In China, DeepSeek’s accomplishment has fueled national pride, with media outlets and tech executives praising the company as a symbol of Chinese innovation in the face of U.S. export controls. In the DW discussion, Winters questioned the long-term impact of chip access and computing power on China’s tech trajectory.