OpenAI and Google have been providing their advanced artificial intelligence (AI) models to Chinese tech corporations that are on the U.S. Ministry of National Defense blacklist.
The Financial Times (FT) reported on the 10th that OpenAI and Google confirmed they provided AI services to Singapore-based subsidiaries of Alibaba, Baidu and Tencent. These are corporations that the U.S. Ministry of National Defense placed on a sanctions "blacklist," judging them to be linked to China's military. FT said, "A gap has emerged in the U.S. government's effort to curb China's AI development."
OpenAI explained that while access to its AI models is not allowed in China, it permits the use of its models by some China-affiliated corporations operating in countries where it can apply safety measures and check whether distillation techniques are used. Distillation refers to a method of creating a model with similar performance by using answers produced by another AI model as training data, and U.S. AI corporations, including Anthropic, have pointed out that Chinese corporations are rapidly improving the performance of their models through this.
OpenAI told FT, "We want AI based on democratic values to be more widely used than AI controlled by authoritarian regimes," adding, "Access rights should not be determined solely by the nationality of corporations."
Google also said its AI services are offered in Hong Kong and Singapore. Instead, usage policies, including a ban on distillation, apply. Google acknowledged that simply blocking AI access in certain countries cannot completely prevent the theft of AI capabilities through distillation. This is because, if they set their minds to it, relevant experts can use sophisticated methods to circumvent such restrictions.
FT reported that some argue the U.S. government should introduce stronger regulations on AI models, similar to how it restricted exports of high-performance semiconductors to China.