According to reports from The New York Times, The Wall Street Journal, Axios, Bloomberg, and other foreign media outlets, the U.S. government said at a closed-door meeting with leading artificial intelligence companies this Tuesday local time that the frontier model cybersecurity review mechanism it announced in June this year currently applies only to closed-source models, while open-weight models will not be affected, at least for now.
People familiar with the matter said that at the meeting, the U.S. government defined the models subject to voluntary review as closed-source models with the most advanced capabilities and national security risks. Nothing in the mechanism should be interpreted as restricting released open-source models.
If an artificial intelligence company chooses to submit a model slated for release to the U.S. government, the model will enter a pre-release review period of up to 30 days. During this period, the model will undergo a joint review by multiple U.S. government agencies, and company employees will be unable to access it.
The U.S. government is not expected to publicly disclose the complete model review mechanism, but the information is not confidential, and companies that choose to participate can obtain the relevant details.