Insights
DeepSeek: AI Under the CCP’s Watch
If you’ve been paying attention to technology news over the past few days then you’ve likely heard about the Chinese-made artificial intelligence model called DeepSeek. Its latest version was released on January 20th, and they claim that it was built at a fraction of the cost of industry-leading models in the United States like ChatGPT, Claude, and Gemini.
Even if DeepSeek delivers comparable performance at a lower cost, is it worth the risk? The Chinese Communist Party (CCP) has long pursued a strategy of using technology to suppress and manipulate information, shaping global narratives to serve its interests.
The reality is that any Chinese tech company operating under government sponsorship (as all Chinese companies do) must comply with strict state regulations on what can and cannot be said. This includes adhering to censorship laws and serving as a conduit for propaganda when required.
For example, the Chinese government's official stance on COVID-19 is that it did not originate in a Wuhan lab—a position that aligns with its broader efforts to control the narrative. However, this claim has been challenged by intelligence agencies, including the CIA, whose assessment suggests that the virus most likely originated from a lab.
If you ask DeepSeek about the origins of COVID-19, it will tell you that the virus “most likely” evolved naturally—a response that conveniently aligns with the CCP’s official narrative. This raises serious concerns about the model’s impartiality and the extent to which it is influenced by state-mandated censorship.
DeepSeek is also applying guardrails on questions about Tibet, Tiananmen Square, and Taiwan—topics that are well-documented areas of Chinese censorship.
For example, Taiwan, a self-governing democracy, is falsely claimed as always being part of China’s territory.
Artificial intelligence is increasingly influencing how people access and understand information, and guardrails are important in making sure these systems are safe, ethical, and reliable. As an AI company dedicated to supporting the criminal justice system, we understand that special accommodations are a necessity in our product TimePilot to ensure that guardrails do not prevent law enforcement from seeing both inculpatory and exculpatory evidence when investigating crimes.
But when a platform like DeepSeek openly displays ideological bias and puts guardrails on critical topics, it becomes a dangerous tool for communist propaganda.
Typically, one might expect this type of censorship to occur behind the scenes, filtering content before it reaches the user. However, that doesn’t appear to be the case with this tool. DeepSeek has automated guardrails seemingly designed to regulate its responses—boldly erasing any points it deems uncomfortable, making its manipulation of information glaringly apparent.
Here’s an example:
The only logical conclusion you can draw is that DeepSeek is designed to be a dumping ground for our intellectual property, targeting those who believe they’re getting a better deal compared to pricier models like ChatGPT.
It’s akin to buying cheap, poorly made toys from China for your kids. Sure, they might save you some money upfront, but when they’re painted with lead and pose serious risks, is it really worth it? You wouldn’t knowingly expose your child to lead poisoning just because it’s cheap—so why take that same risk with your data and intellectual property?
Don’t fall for the trap.
