Sources cited by The Hindu and Channel NewsAsia report that Trump administration advisers have told AI firms they will not require safety testing for “open-weight” AI models. The reporting frames the decision in the context of growing U.S. scrutiny after recent AI-related incidents and hacks that lawmakers say have highlighted potential cybersecurity risks. Lawmakers’ concerns focus on whether increasingly capable open-weight models—whose model parameters or weights can be publicly available or widely accessible—could be used to conduct, facilitate, or otherwise support cyberattacks.
The sources say the advisers communicated this position to companies involved with open-weight model development, but they do not describe detailed alternatives to safety testing or specific regulatory mechanisms that would replace such assessments. The reports add that the debate is occurring as policymakers weigh how to manage the benefits of open AI systems against the risks that may arise from broad accessibility. Both outlets present the information as coming from sources rather than from a formal announcement.
