OpenAI to restrict API access for unsupported countries in July
Key Points
- OpenAI plans to enforce stricter API restrictions for unsupported countries starting July 9, likely affecting China, Russia, North Korea, and Iran. Developers will need to find ways to verify and potentially modify their API usage to avoid penalties.
- A recent report by OpenAI revealed that state-sponsored actors from Russia, China, Iran, and Israel have misused its AI models for propaganda and disinformation campaigns. These attempts had minimal reach and were often detected due to human error, OpenAI said.
- Still, it's an election year and things could get worse, so the API restrictions are part of OpenAI's efforts to prevent such misuse.
Starting July 9th, OpenAI plans to enforce tighter API restrictions on unsupported countries.
While not explicitly named, China (including Hong Kong), Russia, North Korea, and Iran are likely to be the main-affected countries. These nations are notably absent from OpenAI's list of supported countries.
The information originates from a warning email from OpenAI about API usage from unsupported regions to a developer who then tried to comply with the policy by blocking traffic from those regions.
The developer, using multiple API keys for different projects, didn't know how to identify which keys were generating traffic from unsupported regions. OpenAI doesn't seem to provide this information yet.
After repeated inquiries, the support team indicated that the access was from China or Hong Kong, apparently triggered by a Cloudflare worker app (about 5 percent of the traffic). It seems that developers are now tasked with finding their own methods to verify and modify their API usage. to avoid OpenAI penalties.
OpenAI services are being misused for propaganda
In a recent report, OpenAI revealed that state-backed actors from Russia, China, Iran, and Israel have misused its AI models for covert propaganda operations and online disinformation campaigns.
The campaigns included AI-generated content alongside traditional formats, but achieved minimal reach or engagement, according to OpenAI's findings. Human error, such as accidentally posting system messages ("As a large language model..."), often exposed the content as AI-generated.
Earlier in February, OpenAI and Microsoft identified five state threat actors from China, Iran, North Korea, and Russia. They subsequently deleted their accounts for exploiting AI services to conduct malicious cyber activities, including translating technical articles, debugging code, and creating malicious scripts or content for phishing campaigns.
The potential for AI misuse seems enormous, especially in an election year. Making the API more restrictive is part of OpenAI's effort to curb this misuse.
AI News Without the Hype – Curated by Humans
As a THE DECODER subscriber, you get ad-free reading, our weekly AI newsletter, the exclusive "AI Radar" Frontier Report 6× per year, access to comments, and our complete archive.
Subscribe now