Five months after ChatGPT kicked off an artificial intelligence investment spree, Beijing moved to rein in chatbots in China, signaling the government’s determination to maintain tight regulatory control over a technology that could define an era.
China’s Cyberspace Administration of China this month published draft rules for so-called generative artificial intelligence — software systems, such as the one behind ChatGPT, that can generate text and images based on users’ questions and prompts.
Under the rules, companies must abide by the Chinese Communist Party’s strict censorship rules, just as websites and apps must refrain from posting material that discredits China’s leaders or restates banned history. The content of the AI system needs to embody the “core socialist values” and avoid information that undermines “state power” or national unity.
Companies will also have to ensure that the words and images their chatbots create are authentic and respect intellectual property rights, and will be required to register their algorithms, the software brains behind the chatbots, with regulators.
The rules are not final and regulators may continue to revise them, but experts say engineers building AI services in China are already figuring out ways to incorporate the edicts into their products.
Governments around the world have marveled at the power of chatbots, with AI-generated outcomes ranging from worrisome to benign. Artificial intelligence has been used to ace university exams and create a fake picture of Pope Francis in a puffy coat.
ChatGPT, developed by U.S. company OpenAI and backed by about $13 billion from Microsoft, has spurred Silicon Valley to apply the underlying technology to new areas such as video games and advertising. Venture capital firm Sequoia Capital estimates that AI businesses could eventually generate “trillions of dollars” in economic value.
In China, investors and entrepreneurs are racing to catch up. Shares of Chinese artificial intelligence companies soared. Some of China’s largest technology companies have made high-profile announcements, including most recently e-commerce giant Alibaba; facial recognition software SenseTime; and search engine Baidu. At least two startups developing Chinese alternatives to OpenAI technology have raised millions of dollars.
ChatGPT is not available in China. But in the face of a growing number of homegrown alternatives, China has been quick to unveil AI redlines, ahead of other countries still figuring out how to regulate chatbots.
Kendra Schaefer, head of technology policy at Trivium China, a Beijing-based consultancy, said the rules demonstrate China’s “move fast and break things” approach to regulation.
“Because you don’t have a two-party system where both sides argue, they can just say, ‘Well, we know we need to do this and we’ll fix it later,'” she added.
Chatbots are being trained on massive amounts of the internet, and developers are grappling with the inaccuracies and surprising questions they sometimes spit out. On the face of it, Chinese rules require a level of technical control over chatbots that Chinese tech companies have yet to achieve. Even companies like Microsoft are still fine-tuning their chatbots to eliminate unwanted responses. The standards in China are much higher, which is why some chatbots have been shut down, while others are only available to a limited number of users.
Experts are divided on the difficulty of training artificial intelligence systems to consistently match the truth. Some doubt that companies can explain the scope of China’s censorship rules, which are often sweeping and ever-changing, even requiring censorship of specific words and dates, such as June 4, 1989, the day of the Tiananmen Square massacre. Others argue that, with time and enough work, machines can align themselves with truth and a particular value system, or even a political system.
Analysts expect the rules to change after consultations with Chinese technology companies. Regulators can relax enforcement so the rules don’t completely disrupt the technology’s development.
China has a long history of censoring the internet. Throughout the 2000s, the country built the world’s most powerful information net on the Internet. It scares away non-compliant Western companies like Google and Facebook. It employs millions of workers to monitor Internet activity.
All the while, Chinese technology companies that must play by the rules have thrived, defying Western critics’ predictions that political control would cripple growth and innovation. With the advent of technologies like facial recognition and cell phones, companies help countries use them to create surveillance states.
Matt Sheehan, a Chinese artificial intelligence expert and fellow at the Carnegie Endowment for International Peace, said the current wave of artificial intelligence poses new risks for the Communist Party.
Mr Sheehan said the chatbot’s unpredictability to make absurd or false statements – what AI researchers call hallucinations – ran counter to the party’s obsession with managing online speech.
“Generating AI strains the party’s two highest goals: controlling information and leading AI,” he added.
Experts say China’s new rules aren’t all about politics. For example, they aim to protect the privacy and intellectual property of individuals and creators of AI model training data, a topic of global concern.
In February, image database company Getty Images sued AI startup Stable Diffusion for training its image generation system on 12 million watermarked photos that Getty claimed diluted the value of its images.
China is making a broader push to resolve legal issues around the use of underlying data and content by artificial intelligence companies. In March, Beijing established the National Data Administration as part of a major institutional overhaul aimed at better defining what it means to own, buy and sell data. State agencies will also assist companies in building the datasets needed to train such models.
“They’re now deciding what property data is, who has access to and control over that data,” said Ms. Schafer, who has written extensively on China’s AI regulations and called the move “transformative.” .
However, China’s new protective measures may not come in time. The country is facing increasing semiconductor competition and sanctions that threaten to dent its competitiveness in technologies including artificial intelligence.
Expectations were high for China’s AI in early February when AI engineer and entrepreneur Xu Liang released one of China’s first mobile apps, ChatGPT. The app, called ChatYuan, garnered more than 10,000 downloads within the first hour, Mr. Xu said.
Media reports soon surfaced of stark discrepancies between the party line and ChatYuan’s response. The response offered a bleak diagnosis of the Chinese economy and described Russia’s war in Ukraine as a “war of aggression”, at odds with the party’s more pro-Russian stance. Authorities shut down the app a few days later.
Mr Xu said he was adding measures to create a more “patriotic” robot. They include filtering out sensitive keywords and hiring more human reviewers to help him flag questionable answers. He’s even training a separate model that can detect “incorrect opinions,” which he will filter.
Still, it’s unclear when Mr Xu’s robot will satisfy the authorities. According to screenshots, the app was initially slated to come back up on February 13, but was still down as of Friday.
“Service will resume after troubleshooting is complete,” it wrote.
