Tencent’s chief AI scientist, Yao Shunyu has warned that artificial intelligence models could soon pose biological risks similar to the threats they already present in cybersecurity.
“The cyber abilities of [AI] models and soon [the] bio ability of models will reach a threshold where it’s actually possible to do harm,” Yao said, according to the SCMP.
He issued the warning at a science event in Hong Kong on Saturday, according to the South China Morning Post.
Yao urged Chinese technology companies developing open-source AI models, which allow users to download and modify them, to prioritise safety measures.
He also cautioned that managing biological threats posed by AI could prove more challenging than addressing cybersecurity risks.
Yao, 29, joined Tencent last year after spending a year at OpenAI. His comments came days after the Chinese technology company unveiled its new flagship AI model, Hy4, on 4 November.
In its launch announcement, Tencent said it had excluded material on exploiting software vulnerabilities from Hy4’s training data, focusing instead on defensive cybersecurity information.
The company also said Hy4 has a strong foundational understanding of biology but is less capable than leading AI models in designing experiments. Tencent said this assessment was based on internal reviews by experts.
Tencent plans to release Hy4’s model weights and a report detailing its safety tests in the coming days, Reuters reported.
Hy4 is currently ranked as the second-best Chinese AI model, according to benchmarking firm Artificial Analysis, the South China Morning Post reported.
