Hello!
If your news sources and socials are anything like mine, you could be forgiven for forgetting that late last month a landslide in Nepal near the border with Tibet caused a debris flow that wiped a border crossing, Gyirong Port, off the map and killed at least 1,300 people, with thousands more still missing.
In a year when 6,300 people have died from recent earthquakes in Venezuela, 1,800 migrants have been reported dead or missing on the Mediterranean, and 3,000 to 30,000 protesters were killed by the Iranian state (not to mention at least 3,400 Iranians who were killed by US and Israeli attacks), and when perhaps 50,000 civilians have been killed in Gaza by Israeli forces since 2023, it’s easy enough to see these most recent deaths as a tragic accident in a nasty and brutish world and move on.
The Chinese government doesn’t mind if we do. The flow of information in and out of Tibet, which was annexed (or ‘peacefully liberated’) by China 75 years ago, has long been tightly restricted, and the latest disaster is no exception. While the Nepalese government has provided regular updates about the dead, missing, and rescue operations, China has been slow to release information and given low estimates for the number of deaths and missing.
Sensitive to suggestions that it could have provided an advance warning or that development of the region may have contributed to the tragedy, and to criticism over maintaining Gyirong Port in the same location after past natural disasters, the Chinese government has maintained very tight control over information within the country as well.
The Taipei-based China Media Project has published a detailed analysis of the ways the Chinese state controlled media narratives about the catastrophe, looking at both Xi Jinping’s thirteen-year campaign to rein in local news and the specific dynamics of publication over the last few weeks. News outlets waited five hours before publishing anything at all, likely waiting for a state-sanctioned angle. The comparison with a 2008 earthquake, where hundreds of reporters openly defied a ban from the Central Propaganda Department to go to the scene and report, is striking.
Individuals, meanwhile, are punished for ‘spreading rumours online’ or ‘disrupting online order’ – in the case of Tibet, sharing information with journalists or advocates can lead to imprisonment. At least ten people have already been punished by the Cybersecurity Bureau of the Ministry of Public Security as part of a ‘Clean Web 2026’ campaign for posting claims that hundreds or thousands had died, contrary to official statistics.
The US-based China Digital Times determined that the video censorship was based on the images themselves, rather than accompanying text, pointing to use of video fingerprinting or something of that nature. Such automated platform censorship means that even if individuals have potentially objectionable footage, they can’t do much with it, or at least not for long. During the coverage gap, WeChat users tried and failed to share video clips of the destruction, search results were limited, and hundreds of topics related to the events were deleted from Weibo.
These censorship tactics form part of the government’s policy of ‘public opinion guidance’, which aims, in part, to prevent or minimise large-scale interest in any given topic, for fear that mass attention could turn into unpredictable mass criticism. The other side of this tactic is distraction. Social media platform staff report that during crises platforms sometimes choose to promote entertainment stories, redirecting audience attention and thus evading regulatory pressure – a dispute between an actress and a crypto billionaire seems to be playing this role now.
The Chinese government knows intimately the risks of the open internet during natural disasters – read Tom Johansmeyer’s Binding Hook article for more on how states exploit post-disaster chaos.
After the first state news release, which followed a long-established playbook of focusing on ‘positive energy’ and government-led relief efforts, a content farm including seemingly personal accounts – for example, a food blogger and a car-review channel – spent several days parroting the official release with personal, emotional touches.
It seems China has moved away from unsubtle mass deletion campaigns to a strategy of saturating the information environment with safe, visually appealing content, combined with more selective removals and means of preventing information that might prompt criticism from ever being shared.
As anyone on the internet these days is aware, AI is the perfect tool for conducting that saturation campaign. As Human Rights Watch points out, as real videos of the disaster are censored, official channels share AI-generated videos that claim to explain it, notably devoid of the poignant human figures futilely running seen in the originals.
And Anthropic’s latest threat report shows Chinese state actors using Claude to assess social media and news content for political sensitivity and flag targets for ‘control’ (SecDev Group’s Greg Walton has a good summary of the Tibet implications). One of the accounts mentioned in the threat report as being monitored was also a high-profile distributor – on X – of footage from the landslide.
Read Siena Anstis’ Binding Hook article on how gaps in international law and norms facilitate cross-border digital repression.
The Anthropic report obviously only explores activity they’ve uncovered and actors that used Claude and not the multitude of other AI platforms out there – one must presume there are a lot more Chinese state actors using AI for a lot more things.
All in all, this horrific – and potentially climate-change exacerbated – natural disaster serves as still further evidence of the power and effectiveness of Chinese internet controls, even in a country where more than one billion people have portable, pocket-sized devices with cameras and internet connections. It also serves as an additional reminder that AI’s negative impacts are not limited to grandiose world-destroying narratives – wielded by authoritarians, it can contribute to a gradual numbing of genuine emotion and human empathy.
Until next month,
Katharine Khamhaengwong
Binding Hook Senior Editor
More Binding Hook on Chinese information gathering and control:
- Rogier Creemers discusses how China has maintained its grip on the internet amidst the rise of generative AI.
- Paulina Uznańska examines the information-gathering potential of Chinese-made vehicles and how the EU might combat that by following China’s own regulatory path.
- Bernhardt Fourie looks at the ways that compliance with Chinese cybersecurity regulations might allow the state to obtain a bit too much data from multinational corporations.
- Joseph Christian Agbagala argues that the role of provincial governments in Chinese cyber operations, including intelligence gathering, is underexamined.
- Neil Ashdown writes about threats to research security, including from China.
