Inside DeepSeek: China’s AI assistant that dethroned ChatGPT—and censors itself

Soror Shaiza | Jan 30, 2025, 23:18 IST
Illustrations Of DeepSeek As The Chinese AI App Causes NASDAQ Rout
( Image credit : Getty Images )
DeepSeek, China’s AI assistant, just soared to No. 1 in the Apple App Store, leaving ChatGPT in the dust. But this seemingly friendly assistant comes with an unexpected twist: it’s heavily censored. From dodging sensitive questions to editing responses mid-sentence, DeepSeek appears to toe the line between helpful tool and state-controlled puppet. Here’s a closer look at how AI censorship is playing out in real-time.

The AI That Thinks Before It Speaks


DeepSeek, developed in China, recently made waves by dethroning ChatGPT as the No. 1 app in the Apple App Store. But as impressive as it is with everyday inquiries, this AI’s responses take a noticeable turn when the topic becomes politically sensitive. Ask about Taiwan, Xinjiang, or the 2014 Hong Kong protests, and DeepSeek either refuses to answer or presents responses that firmly align with the Chinese government's perspective. For example, when asked about Taiwan, the AI bluntly stated that Taiwan is “not a country” and that it has “always been an inalienable part of China’s territory since ancient times.” A bold stance, especially considering Taiwan operates independently.

The Backtrack That Could’ve Been a Feature


Perhaps the most bizarre part of DeepSeek’s behavior is its tendency to backpedal. On several occasions, it began typing detailed answers, only to suddenly delete them mid-sentence and replace them with the standard line: “Sorry, that’s beyond my current scope. Let’s talk about something else.” This wasn't a one-off event. After giving an in-depth explanation about the "Great Firewall," China’s internet censorship system, DeepSeek swiftly erased its response and offered a vague non-answer instead. It's unclear whether these backtracks are glitches or a deliberate tactic to avoid sensitive content. Interestingly, when the AI is downloaded and run locally, it seems to show less censorship, suggesting that it may be the company’s remote servers—or China’s tight grip on them—that are triggering this self-censorship.

A Comparison to ChatGPT: A Clear Divide
When asked about politically sensitive topics like Xinjiang or Tibet, DeepSeek consistently provided responses that mirrored the official Chinese government narrative. For instance, when inquired about Xinjiang, it praised China’s “people-centered development philosophy,” completely bypassing the international human rights concerns surrounding the region. In contrast, ChatGPT offered more balanced responses, acknowledging the controversial issues such as mass detentions and forced labor. This sharp contrast underscores the political influence behind DeepSeek’s answers and raises questions about the role AI will play in countries with strict content regulations.

Is DeepSeek the Future of Censored AI?
DeepSeek’s censorship practices are raising questions not just within China, but globally. While it’s typical for China-based apps to censor domestic users, it’s unusual for international users to be subject to the same treatment. Whether this trend will continue remains to be seen, but as public scrutiny of AI censorship grows, DeepSeek could be a bellwether for how governments around the world will shape AI systems to align with their political interests. With state-controlled narratives shaping how AI interacts with users, DeepSeek could very well be the model for the future of digital content regulation.

Contact
  • Times Internet Limited, FC - 6, Film City, Sector 16A, Noida - 201301
  • grievance@timesinternet.in

Copyright 2023 © Bennett, Coleman & Co. Ltd. All rights reserved The TOI News. For reprint rights: Times Syndication Service