China has already deployed its expertise in artificial intelligence to erect a surveillance state, power its economy and develop its military. Now Taiwan’s cyber-security chiefs have identified signs that Beijing is using AI to interfere in an overseas election for the first time.
In the run-up to its general and presidential elections on Saturday, Taiwan has detected what appear to be experiments with AI-generated messaging amid disinformation unleashed by Beijing and its proxies. This could presage China’s export of its Orwellian tools for manipulation and control to influence other democracies.
If Chinese programmers can teach intelligent machines to mimic the language of voters — learning idioms, slang and mindsets via elaborate algorithms — it will be a game-changer, spreading fake news and disinformation through anonymous social media accounts at viral speeds.
Cyber-security chiefs believe that the traces of AI messaging in the campaign are tests as China tries to deploy the technology in a more sophisticated manner.
Tzeng said tell-tale signs of AI material were unfinished messages and low-quality wording. “It seems they are trying out new things,” he said. “China uses Taiwan as a laboratory to test its newly developed cyber-warfare tactics for future applications elsewhere.”
Michael Cole, editor of the Taiwan Sentinel, said there was “accumulating evidence that Beijing has begun experimenting with AI to generate false content and disinformation’’.
“We’re seeing the first steps towards using AI and computers to ‘write’ news, using a few keywords, that seems credible,” he said. “We’ve also been seeing evidence of automation in the sharing, almost instantly, of disinformation on social media. I think AI will be the next phase in Beijing’s efforts to overload and saturate the Taiwanese information environment.”
It is “a laboratory for China for adaptation and improvement on political warfare instruments which can then be unleashed against other targeted democratic societies”, he said.
The stakes in Taiwan could hardly be higher during an election that will shape the country’s relations with the US and China. But Tzeng believes that Beijing has more far-reaching goals.
“China has great ambitions for how AI can influence other countries,” he said.
He said China wanted to surpass Russia by not simply interfering in elections, but sowing chaos and confusion and pursuing divide-and-rule tactics in open societies.
A spokesman for Taiwanese President Tsai Ing-wen said: “Beijing’s mission is to try to show that our democracy does not work.”
Even before this election, Taiwan was the democracy most heavily bombarded by false information from abroad, according to researchers at Gothenburg University. Taiwanese officials said the nation’s public sector faced an average of 30 million cross-border cyber attacks a month in 2018, when China was widely believed to have interfered in provincial elections that the KMT won in a landslide.
Messages pushed out by Chinese troll farms often stood out for their poor grasp of the rough and tumble of Taiwanese politics and even for their use of the mainland’s simplified Chinese alphabet (Taiwan uses traditional characters).
China has since employed Malaysian-Chinese and even Taiwanese nationals to generate fake content without the giveaway “linguistic footprint”, Cole said. But a much more ominous achievement would be if China’s programmers and linguists overcame the language barrier to develop a plausible AI “voice”.
Kai Strittmatter, the author of We Have Been Harmonised, which documents China’s use of technology to build its surveillance state, said it would be “an entirely logical next step” for Beijing to use AI in its foreign influence operations.