OpenAI's Chief Financial Officer, Sarah Friar, has raised concerns about the cybersecurity risks associated with adopting low-cost Chinese artificial intelligence models. During a speech at the International AI Summit in Dublin, she emphasized that businesses should prioritize Western models for greater security. Friar highlighted that Chinese models, while often open-source or open-weight, pose a risk as they provide malicious actors with the tools to exploit vulnerabilities. This warning comes amid growing competition from Chinese labs, which have released advanced open-weight systems at lower costs than their U.S. counterparts.
The comments align with broader concerns from companies like Anthropic, which have warned about industrial-scale distillation campaigns where Chinese developers extract reasoning from closed U.S. models to accelerate their own systems. Anthropic recently reported a significant reduction in such extraction efforts, citing Z.ai's activity in developing its GLM-5.3 model. OpenAI, too, has delayed the release of its next-generation model, citing the need for further safety and alignment work. The company has also slashed token prices in response to the influx of Chinese open models, sparking market volatility.
The debate over open-weight models extends beyond technical concerns to geopolitical and commercial strategies. Anthropic's CEO, Dario Amodei, has framed Chinese AI progress as a national-security issue, warning of scenarios where advanced agents could enable large-scale cyber damage or outpace oversight. While he advocates for safety testing and restrictions on advanced chips, critics argue that these measures also serve to protect market position. Meanwhile, former Google CEO Eric Schmidt has warned of the potential for kinetic strikes on data centers if nations fall behind in the AI race, highlighting the growing risks of an unchecked AI arms race.