New AI Nat'l Security Memo Coming This Summer, US Officials Say
The Biden administration is close to issuing a new national security memorandum on artificial intelligence, which is expected to address technology security issues surrounding advanced AI models and related software, National Security Council officials said this week.
Sign up for a free preview to unlock the rest of this article
Communications Daily is required reading for senior executives at top telecom corporations, law firms, lobbying organizations, associations and government agencies (including the FCC). Join them today!
The memorandum should be published “soon,” Maher Bitar, the NSC’s coordinator for intelligence and defense policy, said during a June 24 event hosted by the Council on Foreign Relations. “I think our lawyers told us that we can say this summer.”
The memorandum, mandated by an executive order on AI signed by Biden in October (see 2310300029), could lead to more “standardized, uniform” AI guidance across the various agencies regulating the technology, Bitar said. Part of that includes helping U.S. research institutions and companies manage the risks caused by foreign “adversaries” trying to acquire advanced AI software.
“These are companies in the United States that are making rapid advances, and they are a significant target both for competitive reasons and also by foreign adversaries,” he said. “It's going to be an area of work that we're going to build into the national security memorandum.”
Asked specifically whether the U.S. is considering new export controls on advanced AI models, NSC official Tarun Chhabra didn’t give a clear answer. “We are keeping a close eye on the latest models as they are released,” said Chhabra, the NSC’s senior director for technology and national security. He said the U.S. has been studying both open source and “proprietary” AI models to better understand their “capabilities” and determine how they could be used by a foreign country in a way that threatens U.S. national security.
Although the officials acknowledged that export controls have a role to play, they said they are focused on helping labs and companies improve their cybersecurity to prevent theft of AI-related technology. “That's an ongoing conversation that we have with the leading labs that are doing that,” Chhabra said. “I think there's a broad agreement in the government and with those firms that we need to do more in that space.”
Bitar made similar comments, saying the upcoming AI memorandum, which is due to the president next month, will help answer how U.S. agencies can better “protect U.S. national security interests from foreign use and misuse of technology.”
“We are driving very hard to get this done in time,” he said. “We have every expectation that we will, but it's a complex task.”
Chhabra was also asked about export controls on semiconductors and whether the current U.S. chip restrictions (see 2310170055) are incentivizing Beijing to develop its own, domestic advanced chips and chip technology. When it was deciding whether to impose the restrictions, he said, the administration analyzed sectors where the U.S. hadn’t imposed export controls, such as in the solar and telecommunications industries. “In general," he said, "I don't think that's worked out well for the United States or for our allies.”
“We had to make a decision in the early days of this administration about whether we were going to allow China to continue on the indigenization drive while having full access to all this technology in the meantime,” Chhabra said. “We were committed to ensure that that didn't happen for the semiconductor sector.”