New rules aim to check AI before it goes public

The White House just set up a quick review system for the most powerful AI tools before they hit the market. Developers can choose to join in, but they won't have to. The whole process gives the government just 30 days to flag any national security worries. That’s a short window for an industry that moves at lightning speed. Back in May, a similar plan was scrapped because the president thought it might slow down American tech. Tech leaders were ready to celebrate that version in the Oval Office, but the event never happened. Now, the updated order keeps the process voluntary instead of making it mandatory. The government says it’s not trying to control every new AI system, just the cutting-edge ones that could affect cybersecurity.Experts are split on whether this is a good idea. One policy watcher from a free-market think tank called it a decent first move but worries the rules leave too much power in the hands of the NSA director. Giving one agency so much control could lead to misuse, especially when companies and the government don’t always see eye to eye. The push for oversight started after a top AI model showed an unsettling skill: finding weaknesses in everyday software. The company behind it restricted access to only a handful of partners, but the government still called an urgent meeting with bank bosses to talk about the risks. Since then, more organizations have gotten a peek behind the curtain. Some tech companies have praised the plan, saying it’s better than nothing. Others point out the government only realized it needed these rules after tearing down earlier ones last year. Critics say the new policy feels more like damage control than real preparation for the AI age.