By Borjigin Manda

A Southern Mongolian Voice — Part V
 

An ongoing opinion series exploring Southern Mongolian identity, language, culture, history, and the broader questions of cultural preservation and diversity.

 

Read the Series

Author Bio

Borjigin Manda is the founder and editor of Southern Mongolian News. As a Southern Mongolian, he writes about language rights, cultural preservation, and public policy affecting Southern Mongolia, with a focus on documenting cultural identity and promoting greater understanding of the experiences of Southern Mongolians.

Artificial intelligence is changing how governments and other actors can collect, organize and analyze information about people. A September 2026 report by Anthropic, the company behind Claude, gives a concrete example of how these capabilities are being used in AI-powered surveillance operations that target ethnic minorities, religious communities and political dissidents outside China.

The report, Detecting and Countering Misuse of AI: September 2026, was published on September 10. It covers activity Anthropic disrupted between December 2025 and August 2026 across seven harm areas, from fake dating apps to surveillance systems built to identify and monitor dissidents.

For Southern Mongolians and other minority communities, the report is a warning. It shows how AI-assisted information gathering can intersect with transnational repression and the expanding reach of state power. I read it as a description of methods, and I believe those methods deserve our attention.

Its significance becomes clearer next to two measures that took effect this year: China’s Law on Promoting Ethnic Unity and Progress, and the State Council’s new Regulations on Exit and Entry Administration.

From Scattered Posts to Organized Profiles

Anthropic’s report documents several operations in which China-based or China-aligned actors used Claude to support surveillance and intelligence gathering.

In one case, designated GTG-14010, an actor used Claude to turn material from more than 100 monitored WhatsApp groups and dozens of Telegram channels into structured Chinese-language data, and to profile people who might be vulnerable because of financial stress, family separation and ideological disillusionment. The aim was to track and recruit Uyghurs in Syria, offering payment for information on armed formations, while Claude translated replies in real time. Anthropic judged with low confidence that the operator was a contractor, not a state organ acting directly.

A second case, GTG-14020, involved accounts likely tied to a Chinese government-aligned intelligence operation that built Chinese-language dossiers on religious leaders and diaspora figures, and told the model to adopt “China’s standpoint.” The targets included Tibetan Buddhists, Catholic leaders, Taiwanese Christian organizations and Falun Gong practitioners and affiliated media. The operator asked for each target’s China-related activities, “negative information” and vulnerabilities that could serve as leverage.

A third group of cases involved accounts linked to Chinese public security and state security bodies. They used Claude to monitor politically sensitive content, profile overseas dissidents and civil society groups, and prepare recurring intelligence-style briefings. One request sought the gathering point, route and endpoint of a pro-democracy march in Vancouver, and others concerned venues for Uyghur cultural events in Turkey.

None of these is a lone analyst running a few searches. They involve templates, scheduled reports and repeatable workflows. That is why AI-assisted surveillance of overseas communities can no longer be treated as a hypothetical concern.

Why This Matters for Southern Mongolians

The cases documented by Anthropic involve different ethnic, religious and political communities, including Uyghurs, Tibetans and others. What connects them is not a shared religion or geographic location, but the political sensitivity their identities and activities hold for Beijing, particularly when their voices reach audiences beyond China’s borders.

Southern Mongolians share that position. Many of us see Beijing’s policies as a deliberate, step-by-step effort to eliminate our language and script, our culture, and ultimately our existence as a distinct people. Southern Mongolians have long spoken out against this, and the 2020 protests over Mongolian language education showed how quickly the struggle becomes political. Because we speak out, Southern Mongolian activists are exactly the kind of people who become targets of surveillance.

AI is changing the scale of information analysis. It can process material across languages and platforms, connect otherwise separate pieces of public information, and generate profiles with less specialized human effort than before. A photograph, interview, event announcement or organizational record may reveal little on its own. Analyzed together, such material can expose relationships, patterns of activity and networks of association.

When Southern Mongolians protested the 2020 language policy, security officials issued arrest warrants for hundreds of parents, complete with mug shots taken from surveillance cameras. The state already treats our public dissent as a security matter, and already watches it. What AI changes is the cost. Work that once needed teams of linguists and analysts can now be done by one operator, even in a language they do not speak, and the old protection of a small language is weaker than it was.

Anthropic’s report documents the capabilities and the pattern of targets. Southern Mongolians match the profile of its targets: a people Beijing is trying to assimilate, with a diaspora that speaks out. I believe we face the same exposure, and being heard is exactly what makes us visible.

 Technology Meets Law: The Reach of the State

Surveillance matters because of what a state can do with what it learns. This year, Beijing has put two measures in place that widen what it can do.

The Law on Promoting Ethnic Unity and Progress was adopted on March 12 and took effect on July 1. Chinese authorities present it as a framework for unity and common development. Amnesty International has warned that it could be used to justify transnational repression, and that peaceful minority rights advocacy by anyone, anywhere, could be characterized as undermining “ethnic unity.”

The Regulations on Exit and Entry Administration took effect on September 15. They allow exit bans of six months to three years on Chinese citizens whose overseas activities are judged to endanger national security or interests. Human Rights Watch notes that authorities may withhold written notice of a ban on national security grounds, and criticizes the grounds as broad and poorly defined.

I am not suggesting these form one program, and this is my own reading. But side by side they describe a chain. Surveillance shows who is doing what, with whom and where. A broadly worded law decides which of those acts count as a problem. A border rule with wide discretion attaches a consequence with little visible process. The weak link is evidence: a software-built profile, written from the state’s standpoint and read by someone who expects to find a threat, does not need to be accurate to be used. It only needs to look plausible.

A Collective Response: Governments, AI Companies, and Communities

The response must extend beyond individual digital security. No community can independently solve a problem shaped by powerful technologies, state institutions, and cross-border political pressure.

Governments. Democratic governments should treat AI-assisted monitoring of diaspora communities as transnational repression, with the same seriousness as physical intimidation. That means giving security and justice agencies clear mandates to receive and act on reports from minority communities, and protecting lawful overseas protests and cultural events as likely targets of reconnaissance. It means raising the Ethnic Unity Law and the exit-entry rules in bilateral talks and at the UN, where the human rights chief has already urged China to repeal the former. And it means considering sanctions against contractors and agencies that run such operations.

AI companies. Anthropic deserves credit for publishing this, and others should follow. Companies should share findings and indicators with civil society and affected communities, not only with governments and industry peers. Detection should work in minority languages, Mongolian included, and should look at patterns across sessions and repeated attempts, not only single requests. Communities also need a clear channel to report suspected misuse and be heard.

Communities. Human rights organizations, journalists, and diaspora communities should document credible incidents carefully, preserve evidence, and distinguish verified facts from allegations. Cooperation among Southern Mongolian, Uyghur, Tibetan, Hong Kong, and other affected communities can help identify shared methods and bring them to the attention of researchers, media organizations, and policymakers. Such collaboration should strengthen accountability while respecting each community’s distinct experiences.

Conclusion

Anthropic’s report is a partial view, from one company, of one platform. That is a reason to take it seriously, not to relax. Its authors saw this activity because they were looking for it, and much of the world is not.

If surveillance becomes cheaper and the laws around it become broader, how many of the lawful things we do in public will start to feel risky? The answer depends partly on what governments and companies do, and partly on how deliberately our own communities act. We should ask it now, while the answer is still ours to shape.

The question is not whether Southern Mongolians should stop making their voices heard. It is whether the international community will act to ensure that the technologies capable of amplifying those voices are not also turned into tools for monitoring and suppressing them.

Sources

  1.       Anthropic, Detecting and Countering Misuse of AI: September 2026, September 10, 2026.
  2.       Amnesty International, China: New “Ethnic Unity” Law Set to Entrench Assimilation of Minority Groups, June 30, 2026.
  3.       Human Rights Watch, China: New Border Decree Curtails Right to Leave, September 16, 2026.