As OpenAI Blocks China, Developers Scramble to Maintain GPT Access Through VPNs
Despite the restrictions, Chinese developers are accessing OpenAI's AI models using VPNs and third-party services. OpenAI has prevented China from accessing its API, forcing developers to seek alternate access. Microsoft continues to offer AI services in China, but other big AI providers have restricted access.
Despite the restrictions, Chinese developers are finding methods to use OpenAI's AI models using virtual private networks (VPNs) and third-party services that still have access to GPT.
Previously, OpenAI had taken steps to prevent VPN access by blocking IP addresses linked with these services. Chinese developers, on the other hand, have been able to gain access to OpenAI via a Chinese VPN service that still supports ChatGPT. This comes as OpenAI recently limited China's access to its API, marking the first full day of restrictions.
Developers in cities across China, including Guangzhou, Chengdu, and Dalian, have chosen to remain anonymous due to the legal uncertainty surrounding the use of foreign AI technologies and VPNs in the nation. OpenAI has yet to reply to requests for comment on these developments.
OpenAI's move to prohibit access to its API from unsupported locations, including China, Hong Kong, and Macau, is comparable to actions made by other prominent AI services such as Google's Gemini and Anthropic's Claude. These activities are in response to geopolitical uncertainty and the ongoing US-China technology conflict. However, Microsoft's AI services continue to be available in the region.
In China, developers have developed alternative ways to access GPT models, such as by using third-party services like Microsoft Azure. While many Chinese services that provided GPT access were shut down due to regulatory pressure, Microsoft continues to provide corporate access in mainland China and has promised to retaining access to its Copilot AI service in Hong Kong.
OpenAI's ability to properly prohibit unauthorised access to its publicly available services remains unknown. Some developers have discovered ways to hide their connection origins by using foreign mobile phone numbers when connecting to OpenAI's API over VPN. However, there are signs that accessing OpenAI is becoming increasingly difficult in China, where unauthorised VPN use is prohibited and publicly available large language models (LLMs) require licencing. Some developers have reported that their GPT service providers have turned off specific OpenAI services.
Despite these problems, most developers believe that the constraints will have little impact on their work, given there are various LLM alternatives available from both local and global AI firms. Many developers are already creating backup plans, including a switch to Anthropic, whose most recent model, Claude Sonnet 3.5, has received accolades for its capabilities when compared to the most recent GPT-4 versions. However, Anthropic blocks API access in China.
Furthermore, some developers are looking at domestic AI systems that aim to match the capabilities of abroad models. These platforms have been actively looking for ways to offer comparable AI services across the country.
Chinese developers are using VPNs and third-party services to access OpenAI's AI models despite restrictions.
OpenAI has blocked China from accessing its API, prompting developers to find alternative means of access.
Microsoft continues to provide AI services in China, while other major AI services have also restricted access.
Source: SCMP