Qi Xiangdong, chairman of Beijing-based cybersecurity firm Qi An Xin (QAX), told the Digital China Summit in the southeastern city of Fuzhou on Tuesday that large AI models brought security challenges and risks, according to domestic media reports.
“From an external threat perspective, hackers can exploit vulnerabilities or engage in data ‘poisoning’ to manipulate the model’s decisions, committing malicious acts under the guise of a large model,” he said.
“From an internal operations perspective, if the staff involved introduce erroneous information while updating the knowledge base, it can contaminate the model’s learning environment, leading to incorrect outputs.”
Authorities have firmly backed the push for widespread AI use. Beijing has hailed DeepSeek as a success for the country’s innovation drive in the face of Western sanctions that have limited China’s access to hi-tech chips.
