China's DeepSeek: Implications of its Growing Reach in Autocracies and Democracies

Plus new research on spoofed media domains, cultural propaganda, and united front networks in Mexico

For this month’s post I look at how much and where the China-based generative AI model DeepSeek is gaining traction a year after its release, and its implications for Beijing’s foreign influence and local democratic resilience. Also, a few new pieces of research that caught my eye from Graphika and the Jamestown Foundation’s China Brief. Thank you for reading and if you find this analysis helpful, please subscribe and share.

China’s DeepSeek: Implications of its Global Footprint in Democracies and Autocracies

As I was listening to a January 18, 2026 episode of the Wall Street Journal’s “What’s News” podcast a particular detail caught my attention: DeepSeek, a China-based generative artificial intelligence (AI) model, has become the default chatbot on Huawei phones, including in markets outside China.

If true, this phenomenon raises significant concerns about the spread of an AI system known for embedding content restrictions aligned with Chinese Communist Party (CCP) preferences into the responses and coding it offers users—on topics such as persecuted minorities, historical events, or other politically sensitive issues. The potential for such built-in biases to influence not only views of China, but also public discourse, local politics, or even elections in other countries warrants closer examination. I decided to dig deeper into DeepSeek’s global spread. Analysis of recent reporting and data reveals a pattern of adoption that is especially strong in authoritarian regimes, but also extends into parts of the democratic world.

An Associated Press article from January 8, 2026 on Microsoft’s “Global AI Adoption in 2025” report also published that day provides the clearest public reference to the DeepSeek-Huawei link, noting:

“In many places, DeepSeek’s prevalence correlated with it being a default chatbot on widely available phones made by Chinese tech companies like Huawei.”

Microsoft attributes the surge in DeepSeek’s use to open-source availability, zero-cost access, and strategic partnerships—including promotions and integrations with Huawei in regions like Africa. These factors lower barriers in price-sensitive markets where Western alternatives face cost, availability, or regulatory hurdles.

Mapping DeepSeek’s reach in autocracies

The Microsoft report includes a shaded map of DeepSeek’s market share and geographic footprint. Though the report only labels some of the countries, highlighting key authoritarian regimes aligned with China, additional analysis of the map yields further insights—including on DeepSeek’s spread in freer and more competitive political environments.

A map of DeepSeek’s market share around the world, per Microsoft report on global AI adoption. Credit: Microsoft

According to the above map and report, DeepSeek’s highest market concentrations are in China (89%) and countries closely aligned with Beijing or restricted from access to Western tech: Belarus (56%), Cuba (49%), Russia (43%), Iran (25%), and Syria (23%). All of these countries are rated “Not Free” in the latest edition of Freedom House’s Freedom in the World democracy index. But they are not alone. Other nations in the 21-25% range include Afghanistan and Myanmar. In the 16-20% band fall Ethiopia (18%), Zimbabwe (17%), Yemen, Iraq, and others—all rated “Not Free.” These patterns align with areas of elevated Huawei smartphone usage, strong infrastructure presence, and/or limited Western AI access.

Lower but spreading adoption in democracies

DeepSeek’s penetration is not confined to autocracies, though the market share is not as high as in the above list of countries. In the 11-15% range, several developing democracies or hybrid regimes appear: “Partly Free” countries such as Kenya, Zambia, Malawi, and Mozambique join “Not Free” states like Uganda, Niger, Tanzania, Angola, Venezuela, and Laos. Notably, South Africa, classified as “Free” by Freedom House, also falls here. In general, Sub-Saharan Africa features prominently in the AI model’s spread, according to the Microsoft report. The company estimates DeepSeek usage in the region at two to four times higher than global averages, linked in part to Huawei’s share of the mobile device market, which is at least 6 percent.

Lower but still meaningful penetration levels of DeepSeek (6-10%) extend to an even broader and more geographically diverse set of countries, including more established democracies. Examples of countries rated “Free” by Freedom House include Ghana, Botswana, Namibia, Spain, and Canada. “Partly Free” countries encompass Nigeria, Senegal, Indonesia, the Philippines, Bangladesh, Pakistan, Jordan, Colombia, and Peru. (A few “Not Free” cases like Vietnam, Algeria, and Rwanda also appear in that range.)

In total, of 57 countries listed on the Microsoft map as having at least 6% DeepSeek penetration, almost half (26 countries) are rated “Free” or “Partly Free” by Freedom House.

In Europe and North America, overall DeepSeek uptake remains low due to strong competition from U.S. models (Gemini, ChatGPT, Claude), GDPR compliance, regulator scrutiny, and government-level restrictions in several countries.

Interestingly, other Chinese smartphone brands are being more cautious when it comes to their default AI model. Xiaomi and Oppo, which hold much larger global market shares than Huawei, appear to be prioritizing Google’s Gemini for their international devices. Within China, these brands incorporate DeepSeek for enhanced local functionalities, but favor U.S.-owned AI models for their global customers.

Implications and closing thoughts

As it spreads globally, DeepSeek’s documented content filtering and manipulated responses—on Uyghur detentions, Hong Kong protests, Taiwan, Falun Gong, or other issues deemed politically sensitive by Beijing—could subtly normalize CCP-aligned perspectives in everyday interactions, if made accessible to users without retraining. In democracies facing upcoming elections, like Kenya or Peru, a default version of DeepSeek embedded on affordable devices opens the door for Beijing-aligned views to influence public opinion on China-related matters or domestic debates. That has potential implications not only for geopolitical competition, but also for local democratic governance.

So, what can be done? For democratic governments in the Global South, proactive steps to enhance resilience could include auditing pre-installed AI on imported devices or promoting less censored alternatives through incentives. Civil society organizations and researchers should conduct periodic testing and monitor for any manipulation patterns, especially during politically sensitive periods, like the run-up to elections. For analysts of authoritarian influence and alliances, DeepSeek’s grassroots penetration into major autocracies complements those countries’ elite alignment with Beijing, but with the potential to shape broader public opinion.

DeepSeek’s rise captures genuine innovation in China’s tech sector and illustrates how affordability and accessibility can accelerate technology diffusion. But it also highlights potential vulnerabilities for privacy and free expression, while demonstrating how pre-existing global footholds by China-based companies like Huawei can accelerate adoption of new Beijing-favored tools. As AI-driven technology becomes ubiquitous in the coming years, this is unlikely to be the last example of such piggy-backing.

[For those interested, at the end of this post is the full table of 57 countries with at least 6% DeepSeek penetration, per the Microsoft map, and their level of freedom per Freedom in the World 2025.]

• Update: On February 10, the Estonian Foreign Intelligence Service published its 2026 report on “International Security and Estonia.” It included a section on DeepSeek and the results of its own checks for manipulated responses on a range of topics with national security implications for the country. The study found that on questions related to Estonia, as the time frame moved closer to the present, responses became “increasingly ideological, evasive, and opaque” and that DeepSeek’s distortions were especially evident when discussing Russia’s invasion of Ukraine. The service concluded “When discussing issues related to Estonia’s security, DeepSeek conceals key information and inserts Chinese propaganda into its answers”. That full section of the report is available here.

What I’m Reading

• New Graphika report “Glass Onion”: On January 13, the social media intelligence firm Graphika published its latest investigation into a China-linked online influence network. What stood out in this case was the use of spoofed domains and websites designed to impersonate credible international news outlets such as the Wall Street Journal and The Guardian. The company identified 48 domains and 37 subdomains pushing pro-China content, including material from Chinese state media and other CCP-linked entities. Notably, “The domains copied specific design elements directly from the websites of legitimate outlets in an effort to make their impersonation more convincing.” One of the most revealing aspects is the well-documented role of Chinese marketing and PR firms acting as intermediaries: they placed content on the spoofed sites and then promoted the links on Chinese platforms, falsely claiming that prominent international outlets had published favorable stories about their clients. Globally, accounts tied to the previously documented Spamouflage network of fake profiles also amplified material from these fake domains.

• Beijing Dance Academy’s CCP links: Last month’s edition of the Jamestown Foundation’s China Brief featured a detailed article on the Beijing Dance Academy (BDA), its deep control by the CCP, and its deployment by the regime for international cultural propaganda and potential censorship leverage. Drawing on BDA’s own website and other sources, the piece demonstrates that this leading Chinese arts institution operates as a state-owned and state-funded entity “deeply entwined” with the Party. The majority of its leadership consists of CCP cadres, and the school maintains its own propaganda and United Front Work Department personnel. As is common in China, these ties are obscured or omitted in foreign-facing English-language materials. BDA’s curriculum includes mandatory political education, and proof of political loyalty is an explicit eligibility requirement for participation in international exchanges. Beyond the obvious propaganda value to the CCP, the article highlights the risk of globalizing Beijing’s restrictions on artistic expression—citing, for example, a New York university that canceled a contract for Shen Yun Performing Arts shortly after signing an exchange agreement with BDA, despite previously hosting Shen Yun’s shows for over a decade. It is refreshing to see this level of granular analysis on the politicization of a seemingly benign cultural institution whose Western partners are likely unaware of its ideological entanglements.

• United Front presence in Mexico: Tracking CCP influence networks around the world is painstaking work. In another nod to the China Brief, its latest issue includes one of the most in-depth pieces I have seen on this phenomenon in Mexico, showing how the united front system operates in a country with a relatively small Chinese diaspora. Key businessmen—some with ties to the Chinese People’s Political Consultative Conference (CPPCC)—serve as critical vectors for channeling CCP narratives and preferred policies to local officials, and in some cases into the United States. Researcher Cheryl Yu mapped 81 organizations in Mexico linked to the united front system, distributed across the country. Their activities have included issuing statements on U.S. policy toward Taiwan and organizing junkets for Mexican officials to China, which have resulted in bilateral friendship city agreements.

As promised, here’s the full table of DeepSeek penetration per the Microsoft map and Freedom in the World 2025 ratings.