Nvidia has launched Nemotron 3.5 Lightning, a customizable open AI model for high-volume autonomous-agent workloads, as Meta separately released Muse Glimmer and said it will open the weights for Muse Spark 1.2, underscoring a wider U.S. push to compete with Chinese open-weight AI labs. More than 20 U.S. technology companies recently urged policymakers to avoid "premature restrictions" on open-weight models, arguing they can strengthen competition and American technological leadership. Nvidia said Nemotron 3.5 Lightning stems from its Nemotron 3 family released in December and described it as "truly open source" because it publishes training datasets, techniques and model weights for developers to inspect. Meta said Muse Glimmer is part of a strategy to return some of its most powerful AI models to the open ecosystem after developer disappointment with Llama 4 in April 2025 and the company's subsequent shift toward proprietary Muse-branded models. The debate over open and open-weight AI has intensified amid concerns in Washington and Silicon Valley over national security risks from Chinese models and over distillation, a training technique critics say can amount to intellectual property theft. Supporters, including Box CEO Aaron Levie and Forrester analyst Charlie Dai, say domestic open-weight models could give enterprises and government users an alternative to Chinese offerings while improving transparency, customization and deployment flexibility. Skeptics such as Uniphore CEO Umesh Sachdev say Meta must rebuild developer trust after moving away from earlier open-weight releases.