OpenAI Chief Financial Officer Sarah Friar has cautioned companies to carefully weigh cybersecurity risks before utilizing Chinese artificial intelligence models. Her warning arrives as organizations increasingly adopt affordable AI solutions originating from China, attracted by their strong capabilities and the adaptability of open-weight frameworks.
During an appearance on Bloomberg Television at the International AI Summit in Dublin on Wednesday, October 14, 2026, Friar emphasized that enterprises must evaluate the safety and security ramifications of their technology selections. She indicated that Western models might be the better option when cybersecurity is a major concern, though she stopped short of labeling all Chinese models as inherently unsafe.
Why Chinese AI Models Are Gaining Ground
Chinese AI developers have emerged as formidable rivals to United States technology firms by delivering affordable models capable of performing complex tasks like software development and data analysis. Furthermore, certain open-weight systems allow firms to download and operate the software on private servers, granting developers greater autonomy over system utilization.
This adaptability has driven popularity among commercial users and software creators. Nevertheless, the associated security risks rely heavily on deployment methods, data accessibility, and internal protective measures. For enterprises managing confidential consumer data, financial documents, or trade secrets, these factors carry as much weight as model cost and capability.
OpenAI Faces Pressure to Keep Up
Friar’s advisory coincides with escalating market pressure on US artificial intelligence enterprises from their Chinese counterparts. She maintained that domestic firms need to drive technological progress forward while maintaining a strict focus on safety.
According to Friar, OpenAI has previously delayed the release of specific models to conduct thorough safety evaluations. However, over-delaying progress risks skewing the global distribution of advanced AI capabilities, including technologies that could introduce severe cybersecurity threats. The core dilemma for creators is maintaining a velocity that fosters innovation without ignoring potential vulnerabilities.
Security Concerns Extend Beyond OpenAI
Competitor Anthropic has similarly voiced apprehension regarding the hazards tied to open-access AI frameworks. Speaking at the same Dublin conference, Anthropic’s head of international affairs stressed that open models demand rigorous safety evaluations.
Ultimately, a model’s country of origin is only one piece of the puzzle for commercial users. Risk levels are heavily dictated by data management practices, server hosting setups, permission controls, and regular vulnerability testing. While hosting a model on local infrastructure provides enhanced oversight, it does not guarantee a secure implementation by default.
Also Read: OpenAI Safety Alert: 3 Fired Researchers Urge AI Monitoring
What Businesses Should Consider
Organizations adopting AI solutions must establish explicit policies defining approved models for staff use alongside restrictions on shared data. Conducting thorough security audits and restricting access to vital systems will help enterprises mitigate potential exposures.
As the technological race between American and Chinese AI providers intensifies, businesses must navigate their choices by balancing operational expenses against technical performance.
