
China's Biggest Cyber Defender Says AI Has Cut the Cost of Not Patching

Qi An Xin, one of China's largest cybersecurity firms, says AI has multiplied the cost of leaving known software flaws unfixed by hundreds or even thousands of times, and that the same technology is now the main tool it uses to defend its customers. The company's vice president, Chen Huaping, put the figure to a conference in Xiamen organised by the country's public security and cyberspace regulators.

The trigger was a run of vulnerabilities found in PHP, the language behind an estimated 70 per cent of websites. Late last month an Israeli-American cloud security firm used current AI models to discover critical flaws in it, setting off a surge in hacking attempts, and Chinese authorities responded with the first public vulnerability bulletins from the national emergency response system since 2021.
And the exposure was wide. Chen said the number of affected assets in China ran into the millions, which is the kind of number that explains why the state treats AI assisted exploitation as an infrastructure problem rather than a policing one.
The company's reach is what makes its view worth hearing. Qi An Xin was founded in 2014, employs more than 5,000 people and became part of a national team backed by the state after China Electronics invested about 3.73 billion yuan, or US$559 million, in 2019. It counts some 400,000 major organisations as customers, including more than 98 per cent of central government departments, centrally run state enterprises and major banks.
Its answer is an all in AI strategy run by Gong Yushan, who heads the subsidiary handling AI security. He describes two jobs, using AI to defend customers and securing the AI systems those customers run, and he is blunt about the balance of power.
And the numbers behind that judgement are stark. Gong said AI has lowered the skill attackers need and raised the operational demands on defenders, and his threat intelligence team reported that a new group it calls UTG Q 022 broke into Chinese routers and VPN equipment and stole design drawings from a shipbuilding organisation, with malware that investigators concluded was probably generated by AI.
Speed is the product claim. The company's Aisoc platform, built on its own QAX GPT model, assesses a security alert in about nine seconds against roughly five minutes for a human analyst, and at one provincial power grid customer it handled more than 10,000 alerts a day where a specialist manages a few hundred.
Qi An Xin says AI has made unpatched flaws hundreds or thousands of times more costly
China had millions of assets exposed by the PHP vulnerabilities
The firm became part of a state-backed national team after a 3.73 billion yuan investment in 2019
Its Aisoc platform triages an alert in about nine seconds
Chinese researchers found a jailbroken DeepSeek model used against a dozen organisations
Source: SCMP


