SAN FRANCISCO, July 24 (Xinhua) -- Thirty-two technology companies and industry organizations Friday called on U.S. policymakers to support open-weight artificial intelligence models and avoid premature restrictions.
In a joint statement titled "Open Weights and American AI Leadership," they argued that U.S. leadership in AI depends on building an open ecosystem across all sectors, rather than on a single frontier model.
Open-weight models are AI models that anyone can download, inspect, modify and run on their own infrastructure. According to the statement, they make advanced AI systems more accessible, adaptable and widely available. They also allow organizations to match models to particular tasks and costs, reserving frontier-scale capabilities for the most demanding problems while using efficient and specialized models for other purposes, it said.
Open weights also strengthen competition among model developers and across cloud computing, chips, applications and services, according to the statement. Such competition could encourage innovation, reduce costs and prevent the economic benefits of AI from becoming concentrated in a small number of hands.
The statement said open-weight models also give users greater control by reducing the risk of dependence on a single provider. They allow organizations to control their own data, evaluate and adapt models to meet their needs, and deploy them according to their business requirements.
The statement acknowledged that open weights carry "real and distinct risks." Once released, model weights are beyond the original developer's control, while modified versions are difficult to trace or reverse. However, it said the appropriate response was not to prohibit open weights. As cyberattackers increasingly use advanced AI, defenders also need access to models with comparable capabilities to detect, simulate and respond to emerging threats.
Open models could expand defensive capabilities, increase transparency and allow vulnerabilities to be identified and remedied by multiple teams, the statement said.
It added that relying solely on closed models was not inherently safe. Concentrating advanced AI capabilities in a few closed models could create single points of failure, weaken competition and leave critical technologies in the hands of a few providers.
Open-weight models, by contrast, allow a broader community of researchers and developers to examine model behavior, identify vulnerabilities, develop safeguards and improve systems over time, it said.
On model distillation, the statement said that using the outputs of one model to help train, improve, evaluate or validate another was a widely used model-development technique.
Policymakers should not conflate such legitimate practices with unlawful efforts to extract value from closed models, it said, adding that concerns over unlawful conduct should be addressed through targeted legal and commercial frameworks rather than sweeping restrictions on techniques important to AI innovation.
The signatories included Nvidia, Microsoft, IBM, Meta, OpenAI, Cisco, Dell Technologies, GitHub, Hugging Face, Mozilla, the Linux Foundation, Palantir, and Y Combinator. ■