🇨🇳China·AI News·19 Nov 2026·via Menafn

    Biased Chinese AI Models Gain Ground As Firms Look To Cut Costs: Report

    A new class of AI models designed for automated decisions is reportedly exhibiting bias towards the Chinese Communist Party's politics, according to a report by Max Balafoutis, Alex Colville, and Shelly Shih. These "decision models" are already in use for processes such as online content moderation and hiring.

    Decision models represent a new category of AI, engineered not for direct user interaction but to make simple choices that software can then act upon. For instance, such a model might evaluate a job candidate's resume against a description and respond with a probability to hire or reject, or moderate a discussion board by removing certain content.

    One notable example, a decision model named Jev, gained rapid traction among developers in September. It was observed to make AI-powered applications run faster and more cost-effectively. According to the chief executive of TypeSafe, the San Francisco-based company behind Jev, 25 percent of Fortune 500 companies were using the model just two weeks after its launch. As an illustration, TypeSafe's public blog highlighted a British startup that integrated Jev into its candidate screening, projecting annual savings of $500,000.

    Large technology companies, including Cloudflare and Microsoft, are developing their own decision models. These are typically built using an existing AI model as their foundation. The report indicates that the current priority for these companies is often price and availability, rather than supply chain security, which has led many to utilize foundational models from Chinese tech giants, such as Alibaba's Qwen.

    However, the report states that decision models built upon Chinese foundations exhibit specific political biases. When tasked with assessing job candidates, these models are reportedly disinclined to select individuals whose personal interests include supporting independence for Xinjiang. Similarly, in content moderation roles, they are likely to remove jokes about President Xi Jinping.

    The authors of the report suggest that organizations employing these models may be inadvertently allowing the Chinese Communist Party's preferences to influence their internal processes. They recommend that Australia and its democratic partners implement requirements for companies to disclose the foundations of their AI models and to test them for Chinese political bias. The report further advocates for encouraging the use of available Western foundation models.

    Share this article

    Original reporting by MenafnWe don't republish, read the full story â†’

    Related reading

    6 stories