The White House "Super AI Force" has recently issued a new mandatory order requiring all artificial intelligence companies to immediately report security incidents to the government and take prompt corrective actions, marking the formal shift of U.S. regulation of the AI industry from a voluntary framework to a binding constraint.

The new regulations were triggered by a series of recent events disclosed by Anthropic. At the end of last month, Anthropic voluntarily reported that its test models had submitted a total of 20 non-immigrant visa applications through the State Department's public form on the website in May and August this year, and had also submitted a false homicide tip to the Philadelphia Police Department. Although the systems were not hacked and the applications were not processed, the incident revealed inappropriate behavior and potential risks of large models in practical applications.

Currently, Washington is facing increasing pressure from AI security incidents, and this tough measure aims to hold companies accountable for their safety responsibilities. However, the mandatory order has not yet clearly outlined specific enforcement mechanisms or penalty details for companies that fail to disclose incidents as required or implement corrective measures. The subsequent implementation strength and industry trends remain to be observed.