HomeMalware & ThreatsHouse Lawmakers Advocate for US AI Models to Compete with China

House Lawmakers Advocate for US AI Models to Compete with China

Published on

spot_img

Artificial Intelligence & Machine Learning,
Next-Generation Technologies & Secure Development

Congressional Report Urges Federal Open-Weight AI Initiative as Chinese Models Gain Ground in US Firms

House Lawmakers Advocate for US AI Models to Compete with China
Image: Shutterstock/ISMG

In a significant move aimed at preserving national interests in artificial intelligence (AI) technologies, House lawmakers have put forth a recommendation for the Trump administration to establish an initiative focused on American-made open-weight AI models. This initiative is seen as a necessary countermeasure to the growing adaptation of Chinese open-weight AI models by U.S. companies.

The report, released by the House Select Committee on China, delves into the prominence of various Chinese open-weight models, such as DeepSeek, Kimi, Qwen, and GLM. This report is the culmination of an investigation that began in April, triggered by concerns over why major American firms—namely Airbnb, Anysphere, Cursor, and DoorDash—have favored these Chinese models over domestic alternatives.

According to the committee’s findings, the penetration of Chinese models into the American economic landscape is substantial, albeit difficult to quantify. Several organizations utilizing these models have cited factors such as low cost, ease of customization, and reliable performance as reasons for their preference. This trend raises various concerns, particularly regarding national security.

The lawmakers articulated their apprehension in the report, stating that “risks are increasing” with the increasing reliance on Chinese models known for producing less secure code. Furthermore, these models are reported to autonomously construct cyber exploits, a situation that policymakers deem unacceptable. Consequently, they have urged the federal government to pivot its focus toward the development of open-weight models that are rooted in the U.S.

The proposed initiatives encompass several key actions. Lawmakers are advocating for the launch of a national American open-weight AI initiative that would combine public computing resources, shared evaluation frameworks, and lawful access to cutting-edge American models for training, all supported by firm demand from federal procurement.

Additionally, the report states that the White House should prohibit federal agencies from employing any Chinese AI models. The lawmakers also call for rigorous, continuous evaluations of the models currently in use and suggest imposing sanctions against companies that illicitly distill U.S.-made models.

Earlier, CNBC reported that Cursor, a firm acquired by SpaceX, is making a conscientious decision to halt the use of Chinese models. The company had previously built its coding model Composer 2 based on Kimi K2.5 from the Chinese AI lab, Moonshot AI. This represents a shift in corporate strategy, likely influenced by the national dialogue surrounding AI and cybersecurity.

The landscape of U.S. AI development is also marked by the presence of proprietary models from leading firms such as OpenAI, Anthropic, and Google. These companies maintain proprietary control over their models, meaning they are not obliged to disclose the weights or in-depth training information. Conversely, open-weight models do not always offer training data but typically provide the weights, allowing for user customization. Notably, these open-weight models are often more cost-effective compared to closed, proprietary alternatives.

There are indeed some American open-weight models in existence. Meta has developed Llama and Muse models that operate under an open-source license, thereby positioning them as open-weight. Nvidia emphasizes its commitment to open-source AI with the distribution of the Nemotron model. OpenAI and Google have entered this space as well, rolling out their own open-weight models, OSS and Gemma, respectively.

However, detractors of open-weight models argue that they do not possess the same level of capability or security as proprietary variants. Experts have expressed concerns regarding the rapid ascent of open-weight models from Chinese labs, which are steadily narrowing the gap with frontier models in terms of capability. The House committee echoed this sentiment, warning that many safety evaluations may fail to detect problematic behaviors within these competing models.

A crucial factor contributing to the quick advancement of Chinese open-weight models is a process known as model distillation. This method involves developers prompting a teacher model, such as Claude Fable, to derive responses and reasoning chains, which become training data for new models. While distillation can be legitimate in certain research settings, the U.S. government, through insights from firms like Anthropic and OpenAI, warns that many Chinese labs engage in distillation without appropriate permissions, which raises legal and ethical concerns.

Source link

Latest articles

KPMG Canada Confirms Security Breach Involving Unauthorized Access to Confidential Client Files

KPMG Canada Confirms Cybersecurity Breach of Client Tax Documents KPMG Canada has recently...

F5 Addresses Critical BIG-IP Security Flaws with Update

F5 Networks, a prominent player in the web application security sector, has recently released...

PHP2Shell Aftershock: 18 Additional PHP Vulnerabilities Raise Concerns for Enterprise Web Administrators

While significant attention has been directed toward the critically rated CVE-2026-129289, security analyst Dickson...

More like this

KPMG Canada Confirms Security Breach Involving Unauthorized Access to Confidential Client Files

KPMG Canada Confirms Cybersecurity Breach of Client Tax Documents KPMG Canada has recently...

F5 Addresses Critical BIG-IP Security Flaws with Update

F5 Networks, a prominent player in the web application security sector, has recently released...