AI advance prompting security safeguards

As China accelerates its artificial intelligence push amid the rise of homegrown large language models, cybersecurity risks tied to AI adoption are escalating, underscoring the need for enhanced safeguards to mitigate emerging threats, said Qi Xiangdong, chairman of cybersecurity company Qi-Anxin Technology Group.

Qi, who is also a member of the 14th National Committee of the Chinese People's Political Consultative Conference, said AI is reshaping industries at an unprecedented pace, but is also making cyberattacks more sophisticated and harder to detect.

"AI, especially large language models, is not just transforming production and innovation — it's also bringing new challenges and risks," Qi said.

China's AI ecosystem has seen rapid breakthroughs, with Chinese startup DeepSeek's LLM drawing widespread attention since the Spring Festival this year for its disruptive potential.

Kevin Lu, general manager of Commvault China, said the deep integration of AI technology is reshaping the industry landscape, creating both opportunities and challenges.

"The advancement and application of AI technologies have heightened the complexity of threats, while the data environments of enterprises have also evolved," Lu said. "Businesses must adapt to these new external and internal conditions, which increase the difficulty of data protection, rendering traditional measures no longer applicable."

AI-powered cyberattacks can be launched at an unprecedented scale, making traditional security defenses obsolete, Qi added.

For example, he said AI also enhances the ability to fabricate deceptive content, intensifying threats such as deepfake manipulation, cognitive warfare and phishing scams. Attackers can exploit AI-driven face-swapping and voice-cloning technologies to spread false information and carry out online fraud.

As AI models gradually integrate into critical infrastructure, they could become the new battleground for cyber warfare.

"When large language models are embedded in key areas such as smart cities, industrial control systems and digital governance, they could amplify existing security vulnerabilities. A successful attack on an AI model could trigger widespread disruptions, leading to service outages, production halts and data breaches, posing significant security risks," he said.

To counter the potential risks, Qi said it is necessary for policymakers to introduce documents regarding mandatory compliance requirements for LLMs and data security, in a bid to provide clear guidance for enterprises to strengthen cybersecurity in the AI era.

Businesses should also be encouraged to conduct regular "health checks" on network and data security, helping companies identify and address vulnerabilities to achieve continuous security enhancement. And a special fund should be established to accelerate the adoption of AI plus security innovations, fostering collaboration between leading enterprises across various sectors and cybersecurity firms to optimize security protections in emerging fields such as smart cities, smart energy and smart finance.

Qi said he believes security is the foundation of sustainable innovation, and the goal is to strike a balance — advancing AI innovation while ensuring it remains secure and resilient.