Chinese chatbots to be tested amid terrorist bioweapon fears
- Published
Chinese and Arabic chatbots are to be tested by state-backed artificial intelligence (AI) experts amid fears they risk helping terrorists build biological weapons.
The Artificial Intelligence Safety Institute, a newly-formed government body designed to investigate whether AI systems present national security threats, is planning to examine chatbots that use Mandarin, Arabic, Korean and French.
It comes after Chinese, Iranian and North Korean hackers were blocked from using ChatGPT when bots were used to carry out cyber attacks.
Most safety testing carried out by AI researchers and companies before new systems are released tests the models in English.
Researchers have found that in some cases, chatbots are more likely to give dangerous answers or encourage illegal behaviour if asked in a different language than English.
Meanwhile, AI systems being developed in China and the Middle East are likely to receive less scrutiny than English-language chatbots.