The US is facing heightened cyber threats from adversaries, with AI giving hackers new capabilities. The problem is expected to worsen: Anthropic says its new AI model is so good at finding vulnerabilities in digital systems that it can’t be released to the public. While the strategy has a “combative tone,” it doesn’t specifically mention China, Iran, North Korea, and Russia, a Council on Foreign Relations expert wrote last month.

Cairncross said Wednesday that he remains “very worried” about attacks from Iran amid the war, though he noted Tehran has “never stopped trying to push in this regard … They are a perpetual bad actor.”

He also didn’t specify whether cyber issues will be a topic when Trump meets with Chinese leader Xi Jinping in a summit planned for next month.

“I would expect that the safety and security of the American people will be first and foremost,” he said.

Cairncross also said he wants a “clean, long-term extension” of a 2015 law that provides liability protections for companies that voluntarily share information on cyber threats with the government, expressing confidence Congress would act. “The stakes are very high in this regard,” he said.