Trump pivots on AI regulation

President Donald Trump’s administration is signaling a more active approach to artificial intelligence oversight after months of emphasizing speed and deregulation. The change comes as U.S. officials pursue discussions with China on AI safety and as Google used its annual I/O conference on Tuesday to introduce new models and agent-style products aimed at consumers and businesses. The two developments landed on the same day, putting Washington’s policy debate alongside Silicon Valley’s product race. Officials and executives described both tracks in terms of safety, capability and competition. ### What exactly changed in Washington? Fortune reported on May 19 that Trump officials were moving toward a more interventionist AI posture after public backlash against rapid deployment of advanced systems. The report said the administration had opened talks with China on AI safety, a notable step for a White House that had previously rolled back Biden-era AI guardrails, according to Time’s earlier reporting and other accounts of the administration’s first-week policy moves. Treasury Secretary Scott Bessent told CNBC on May 14 that the United States could engage China on AI because “we are in the lead,” and that the two sides were planning a safety protocol. Reuters reported on May 13 that AI would be at the forefront of Trump’s talks with Chinese leader Xi Jinping, though two U.S. officials said substantive commitments were unlikely. ### Why are talks with China part of an AI story now? (time.com) Beijing hosted Trump and Xi last week as officials from both countries discussed areas where strategic rivalry and risk reduction might overlap. CNBC reported before and during the summit that senior U.S. officials were willing to explore “channels of deconfliction” on AI, framing the issue as one of security as well as commercial competition. (cnbc.com) The Council on Foreign Relations said in an analysis published six days ago that any U.S.-China dialogue would likely be narrowly scoped around safety rather than broad technology cooperation. Reuters likewise reported that the talks highlighted AI’s strategic importance but were not expected to produce major binding commitments. ### What did Google launch at I/O? (cnbc.com) Google on May 19 unveiled Gemini 3.5 Flash, Gemini Spark and Gemini Omni at its I/O developer conference. CNBC reported that 3.5 Flash will become the default model for the Gemini app and AI mode in Search globally, while Chief Executive Sundar Pichai said the model delivers frontier-level capability at roughly half, and in some cases close to one-third, the price of comparable systems. (cfr.org) Google’s own product post described Gemini Spark as a “24/7 personal AI agent” designed to manage tasks under a user’s direction. The company said Gemini Omni can turn text, image and video prompts into high-quality video outputs, and that a new “Daily Brief” product will assemble a personalized morning update. ### Why does Google’s event matter to the policy debate? (cnbc.com) Alphabet used I/O to show that product deployment is moving beyond chatbots and into persistent agents that can act on a user’s behalf. AP reported that Google’s new assistant products are intended to perform tasks proactively, while CNBC said the event was also a test of whether Alphabet could persuade Wall Street that its AI push can keep pace with OpenAI and Anthropic. (blog.google) The Verge reported that many of Google’s new features depend on users trusting the company with more personal data. That trust question overlaps directly with the administration’s emerging focus on oversight, especially as officials weigh how to address safety, disclosure and cross-border coordination while companies release more capable systems. That connection is an inference based on the timing and substance of the two developments. (apnews.com) ### What comes next? Google said on May 19 that Gemini 3.5 Flash is rolling out globally as the default in the Gemini app and AI mode in Search, while Spark and other agent features are being introduced in stages. On the policy side, the next concrete markers are any White House executive actions on frontier-model oversight and any public details from U.S.-China safety protocol talks involving Trump, Xi and senior economic officials such as Bessent. (cnbc.com) (theverge.com)