OpenAI Global Affairs

What happens when one of the world's leading mathematicians starts working with AI? "AI allows me to experiment. I will try crazier things," says Fields Medal recipient and UCLA Professor Terence Tao. Following a recent OpenAI Forum conversation on AI and mathematical discovery, Tao sat down with us to reflect on how AI is changing the practice of mathematics, from exploring new ideas to accelerating research and collaboration. OpenAI Chief Research Officer Mark Chen also shares OpenAI's broader goal of building tools that help scientists and researchers achieve breakthroughs at scale. Together, they explore how AI can expand human potential in mathematics and science while preserving the curiosity and creativity at the heart of discovery.

Luke Xing used Codex to build something he had needed for years: a quick way to make sound feel balanced again after a hearing impairment. This May, he prompted a small desktop app that lets him test his hearing and tune computer audio for his impairment. The app plays test tones across frequency ranges, then lets Luke adjust the gain, or volume boost, separately for the left and right channels. It then applies those settings to headphones, speakers, and other devices, so he can recalibrate before a call, meeting, or song. The problem started in 2018, during a soccer game. Luke was playing goalkeeper when he dove for the ball and was kicked at the base of the skull. He blacked out. Afterward, something in his inner ear changed. He progressively lost roughly 70 percent of the hearing in his left ear day to day, though the impairment fluctuates. He also has tinnitus and occasional vertigo. The injury has imposed social costs: Luke had always been an extrovert. But after the injury, crowded rooms and restaurants were a lot harder to deal with. In a circle of people, he can miss what someone on his left was saying. In a brewery or noisy room, he turns his good ear toward them. At events, he looks for quieter corners. In conversation, he often has to ask people to repeat themselves, then explain that he was almost deaf in one ear. Medical referrals have been slow, and guidance uncertain. Doctors have discussed hearing aids, but his hearing changes enough that specialists have given different advice about whether to wait or proceed. He tried dozens of desktop audio tools. None gave him the control he needed. Codex changed how he approached the problem. Luke gave it a paragraph describing what he wanted. In minutes, he had a working app on his Mac. He has since made small tweaks, but the core worked: test, listen, and adjust until the sound was balanced. The first time he used it to listen to hip-hop through headphones, the sound felt close to what he remembered from before the injury. For those few minutes, he forgot about the impairment. Luke studied math and economics at the University of Chicago and was drawn into practical technical work: helping companies turn ideas into systems they could use. He worked first in consulting, then at Twilio, moving between customers, product, and hands-on technical problem solving. At OpenAI, he now does similar work with AI. Luke is careful to say that his tool is not a replacement for an audiologist. It is a personal aid, a proxy, and a way to recalibrate quickly, on demand. Many people live with problems too specific for mass-market software and too urgent to wait for someone else to build. That is what Luke wants more people to know: the people living with a problem now have more power to build around it themselves. No one understands a struggle like the person living with it, and no one else feels quite the same stakes.

How can AI augment finance skills and make critical thinking and analysis even more important? Join us Tuesday for an OpenAI Forum conversation between Jagdeep Singh Bachher, Chief Investment Officer at the University of California and Sarah Friar, CFO of OpenAI, about how AI is changing the future of finance. They’ll discuss where AI can help finance professionals do more, what skills matter in the future of work, and how OpenAI’s finance team is using AI in its own work. This conversation is for students, early-career professionals, finance leaders, and anyone thinking about how AI is changing analytical work. Link to register in comments.

Gov. JB Pritzker just announced that he will sign Illinois’ frontier AI safety bill into law, making the state a key architect of the emerging approach to governing the most powerful AI systems. Even as work continues toward a federal framework, Illinois is showing how states can help move the country toward a common standard We appreciated the chance to engage with Illinois policymakers as SB 315 took shape, and endorsed the bill before the first committee hearing about it. The law goes further than other state-level frontier safety measures by requiring mandatory independent third-party audits. Companies developing the most capable AI models should not only have strong safety practices in place, but be accountable for demonstrating that they are following them This is reverse federalism in action: states moving the country toward a shared framework for frontier AI safety while we continue to advocate for national action, whether through the executive branch or Congress Rather than producing 50 conflicting sets of rules, Illinois and others are essentially creating a national standard in reverse, one state at a time, building the foundation and momentum for federal action California and New York have already begun moving toward a common model, with safety standards we see as a strong first step and mechanisms to align state and federal law. Illinois has now advanced it with a stronger accountability requirement. And if it plays in Peoria, it should play in Washington, too A federal framework remains necessary. Frontier AI raises national security, economic competitiveness, and public safety questions that ultimately require national expertise and national standards. But in the absence of action from Washington, states like California, New York, and now Illinois are providing a path that other states can follow Thank you to the Illinois lawmakers who drafted and supported this bill, and to Gov. JB Pritzker for preparing to sign it into law

In our latest issue of The Prompt, we note how OpenAI has crossed the milestone of more than 1 million public-sector subscriptions or “seats.” These users span every level of government: from federal agencies and the national labs, to states including Alaska, Connecticut, Massachusetts, Minnesota, Pennsylvania, and Tennessee, and cities like Austin and Fresno. A majority of these 1 million seats are being provided at essentially no cost. This milestone reflects something bigger than growth – it signals how AI is becoming core infrastructure that institutions rely on to do their work. Across the country, public servants are using AI to help them work faster, more accurately, and more effectively on behalf of the people they serve. Teams are summarizing complex documents, streamlining administrative processes, drafting communications, analyzing information, and reducing time spent on repetitive tasks. That means more time focused on delivering services, solving problems, and helping people. For more on how public-sector agencies and workers are using AI, be sure to read (and subscribe to) The Prompt: https://lnkd.in/e57Q3iE4

This week, OpenAI sent a letter to Congressional leaders urging the federal government to address America’s strategic compute shortage as a matter of national security and democratic AI leadership. The core argument: frontier AI systems are becoming strategic assets. They can accelerate scientific discovery, strengthen cyber defense, enhance national decision-making, and shape the balance of global power. But those capabilities only matter if the government can actually deploy them when and where they are needed. That requires strategic compute – a dedicated, ready supply of the specialized chips, data centers, energy infrastructure, and technical operators needed to run frontier AI systems securely, reliably, and at scale. The challenge is timing. Frontier models are improving on the scale of months. New compute infrastructure takes much longer to build. That mismatch creates a national-security gap. Strategic compute is the bridge between frontier capability and government adoption. Like the Strategic Petroleum Reserve, it would give the United States assured access to a critical resource when national security and public purpose demand it. For more on our letter to Congressional leaders urging them address strategic compute capacity, read our latest The Prompt newsletter: https://lnkd.in/e57Q3iE4

A 3 hour live show, 5 days a week and no real rundown in the traditional TV sense. So, Tyler Cosgrove of TBPN built something just for them. Tyler took us behind the scenes of TBPN at a recent OpenAI Academy to show how easy it is to build a tool that's bespoke just for their format. The workflow is practical: capture stories from social platforms, route them into a shared database, organize them by topic, and turn them into a run of show. And it was easy enough that anyone could replicate it without any coding background. You can see a demo of his tool in the video below.

Preventing disease outbreaks means understanding how diseases spread and where resources are needed most. That is an incredibly complex modeling problem. Join Geoffrey Mosoti Nyakiongora, AIA, Senior Technical Product Manager of AI at the Gates Foundation, in the OpenAI Forum to hear how AI and computational modeling can help public health teams prepare for the next crisis. Geoffrey will discuss what COVID taught us about modeling during a fast-moving outbreak, and how AI-enabled tools could help policymakers move from reacting to outbreaks toward detecting them earlier. This conversation is for anyone interested in the intersection of AI and global health, and how quantitative decision-making can support progress toward health equity. 📅 Tuesday, May 26 🕒 1:30–2:30pm PDT 👉 Registration linked in the comments

Kevin Delaney is using AI to revolutionize local journalism, one curated story at a time. His journalism start-up, The San Francisco Standard, has an office with an airy floor plan more akin to a tech company than a traditional newsroom. That's by design, since it has embraced AI tools more than most media organizations and in more creative ways. “San Francisco is a world capital, and it should be seen -- and covered -- that way,” says Kevin, a former Wall Street Journal editor and Quartz editor in chief who now leads The Standard. That ambition is driving an experiment in local news: an “AI-first news interface” designed not to replace reporting, but to help readers navigate it in more personal, useful, and dynamic ways. With a grant from the Lenfest Institute, which receives funding from OpenAI, The Standard moved from contract signing in mid-January to launching the first version of its subscriber app in early April. A reader opening the app gets an AI-generated briefing shaped by their interests, location, and reading habits. It feels less like a traditional news homepage than a smart, evolving feed: a morning rundown, local recommendations, stories surfaced because of what a reader follows, and even highly specific interests like “Berkeley restaurants” or “ocean swimming.” The Standard is using AI to turn reporting into interactive modules that let readers go deeper on a story, answer likely follow-up questions, surface notable people in the news, and draw on the publication’s archive and reporting corpus. One module highlighted City Supervisor Bilal Mahmood, whose work The Standard has covered, and included a Q&A on a new bill he sponsored to protect the city’s LGBTQ residents from the effects of out-of-state laws targeting their communities. It also opens the door to multi-modal journalism, starting with audio briefings. The newsroom is building tools that could let reporters file notes, transcripts, photos, audio, and video into its internal system, where editors can approve them to strengthen the knowledge base behind future coverage and reader experiences. Subscribers can already submit news tips through the app, with human editors reviewing them before they enter the newsroom’s knowledge base. Over time, Kevin says, that could evolve into a trust system for community contributors: “You have a 5.0 rating on Uber as a rider. Could we build the same thing?” If a reader repeatedly submits accurate tips, future contributions could carry more weight with editors. The Standard is using ChatGPT, but is not relying on just one AI tool or model. It is testing multiple systems, choosing the best fit for each task. The through line is usefulness, trust, and fidelity to the reporting. Kevin hopes The Standard can become a model for other media start-ups: journalism rooted in real communities, close to the people and places it covers, and powered by an AI-first approach that makes that reporting more accessible and more relevant.

OpenAI’s Education for Countries initiative is expanding, with Singapore joining a growing group of partners working to bring AI into education systems in ways that strengthen teaching, improve learning, and prepare today’s students for the workforce of the future. As the initiative grows, it is creating a playbook for countries that want to incorporate AI into their education systems thoughtfully and at scale. In the post below, OpenAI VP of Education Leah Belsky explains how that work is taking shape.