OpenAI Global Affairs

One of the most common questions people ask right now is whether Codex is meant to replace ChatGPT. At last's week’s OpenAI Forum, Thibault Sottiaux explained why he sees the two as complementary tools, not competing ones. “Anything where it's manipulation of files on your computer, running things, automation, doing things every couple of hours…There isn’t much that an agent can’t do at this point,” Tibo said. “I don’t expect that everyone will move to Codex, but it is a nice complement to ChatGPT.” In "Codex is for Everyone: Why Codex Matters Beyond Code," Tibo showed how people are already using Codex as a general-purpose agent that can connect to information sources, take actions, automate workflows, and handle tasks that used to take far longer manually. He even demoed an app he built to find the best bread in San Francisco. Watch the full conversation in the link in comments.

Sarah Dully (Sarah D.) is Rhode Island’s Teacher of the Year, and she uses ChatGPT to help keep her lessons to her high-school students alive. When the reference in a handout goes stale, when a class needs a fresh way to approach a text, or when her plan doesn’t land as she expects with the students in front of her, Sarah rebuilds the bridge fast with AI. While growing up in New England, Sarah loved school, because her teachers made it feel personal. She currently teaches juniors and seniors English, practical literacy, and true crime, and ChatGPT helps her teach to different students in different ways. Practical literacy gives students support with the reading, writing, and communication they need after graduation. True crime lets her leverage what already hooks them: researching motive, evidence, credibility, the cultural phenomenon that is true crime, and the ethics of the genre. Sarah’s students are close to adulthood. They are thinking about jobs, college, family, and the lives they will have to build. Her students have to keep up with the economy and the culture, so her lessons do, too. She used to have students write 140 character takes on their readings. But her kids aren’t on Twitter any more, and it’s not called Twitter, and it has no character limits. ChatGPT helps her adapt. She uses it to refresh examples, rebuild handouts, draft sentence starters, simplify directions, sharpen rubrics, and make alternate versions of the same assignment for students who need different first steps. In a true crime unit, that might mean turning a student’s fascination with a case into a lesson on sources and evidence. In practical literacy, it might mean taking a workplace writing task and making it clearer by morning. For Pi Day, while her class was reading Catcher in the Rye, she asked ChatGPT for tenth-grade math problems tied to the novel, and called the activity “Catcher in the Pi.” Her students walked in wondering why they were learning math in English class, then found themselves thinking about Holden Caulfield in Central Park from a new angle. For Sarah, ChatGPT is a jumpstart for a blank page. It helps with the labor around the teaching so she can spend more of her attention on the part that made her love school in the first place: seeing the students clearly enough to know what they need before they say it.

“It's not just one part in isolation that delivers AI," Dion Harris of NVIDIA says. That was one of the key takeaways from The Infrastructure Race: American Competitiveness in the AI Era, a recent OpenAI Forum conversation in Washington, D.C. featuring leaders from OpenAI, NVIDIA, and Oracle. As Dion explained, AI leadership increasingly depends on how well countries coordinate across the full stack: energy, chips, data centers, models, transmission, and applications. The discussion explored what it will take to scale AI infrastructure responsibly in the United States, including expanding grid capacity, strengthening domestic supply chains, accelerating permitting timelines, and building long-term resilience into critical systems.

One of the most important parts of building AI infrastructure is listening first. This week, in partnership with Karl Brutsaert, Charissa Huntzinger, and SB Energy, we hosted a community roundtable in Milam County, Texas with local leaders, educators, and economic development partners to talk openly about opportunities, concerns, workforce development, and how these projects can create real long-term local impact. Across the country, communities are asking thoughtful questions concerning data centers regarding power, water, jobs, and economic value. Those conversations matter, and they are most productive when they happen early, often, and with the right partners at the table. The future of AI infrastructure has to be about more than compute. It needs to include strong local partnerships, workforce pathways, and tangible community benefits alongside the facilities themselves. I’m grateful to everyone who joined the conversation and shared their perspective candidly. Excited to continue these roundtables across all of our infrastructure footprint! Ginger Watkins, Kaylen Bushell, CRL, Brian Nevarez

Today in our The Prompt newsletter, we endorsed two pieces of legislation: 1) the bipartisan federal Kids Online Safety Act (KOSA), and 2) Illinois SB 315. KOSA would help create stronger online protections for young social media users through safer default settings, expanded parental controls, and greater accountability for online harms. The path forward on kids safety, however, also requires AI-specific rules. And we believe KOSA is complementary to the work we’re doing at the federal and state level. Young people should be able to benefit from AI in ways that are safe, age-appropriate, and grounded in real-world support, including referrals to crisis resources and parental notifications in serious safety situations. Illinois SB 315, meanwhile, is a frontier AI safety bill that would establish clear requirements around safety practices, transparency, incident reporting, and accountability for the most advanced AI systems. The legislation closely mirrors frontier safety frameworks already advancing in California and New York, which we view as the start of a consistent, nationwide framework. For more on our endorsements -- plus much, much more -- read our The Prompt newsletter: https://lnkd.in/eMkpjuEg

A once-in-a-generation opportunity to build the electricity grid of the future and ensure America has the AI infrastructure it needs to maintain its technological edge. That was the main takeaway from The Infrastructure Race: American Competitiveness in the AI Era, a recent OpenAI Forum event moderated by Erin Hodges with leaders from OpenAI, Oracle, and NVIDIA. AI may feel digital, but the systems that power it are deeply physical: data centers, chips, power plants, transmission lines, and other key pieces of AI infrastructure. As Nick Edwards, OpenAI’s Energy and Power Projects Lead, put it: “We’re building the physical layers of the future.” The challenge is timing, scale and optimization. Nick noted that new generation projects can take 3-7 years to get approved, built, and brought online, while transmission projects can take 7 to 12 years. If America wants AI infrastructure to come online at the speed this moment requires, every part of the system needs to move in sync. The old wisdom was that knowledge is power. In the AI era, the equation is inverted: power is knowledge. The countries that can turn electricity into intelligence most efficiently, reliably and responsibly will define the next era of technological innovation.  That work is already underway through Stargate, including at our site in Abilene, Texas, where OpenAI is working with Oracle and NVIDIA to build cutting-edge AI infrastructure, create thousands of jobs, and support local economic growth. Drew Bryck of Oracle pointed to transmission as one essential piece of the buildout. Solar power from the Southwest, for example, can help meet demand elsewhere during the day, while hydropower from the Pacific Northwest can help balance the system overnight. But that only works if the country has transmission lines to move power where it is needed. Drew said work starts long before construction is visible. Years before the first servers arrive, Oracle begins meeting with local leaders, utilities, and community partners to hear about local needs and concerns, and looks for ways each data center project can leave the community better off than it was before. Dion Harris, Senior Director of HPC and AI Hyperscale Infrastructure Solutions at NVIDIA, said the AI infrastructure stack has to work as an integrated system. “We can’t talk about any one of those things in isolation,” he said, because AI leadership depends on energy, chips, data centers, models, and applications all advancing together. The panelists emphasized that American AI leadership will depend not only on building better models, but on building the physical infrastructure that makes those models possible, and doing it in ways that benefit communities rather than burden them. As Dion said, “With the proper investment and the responsible investment, I think we’ll see that payoff.”

Codex started as a tool for developers, but has become a practical AI tool for anyone. Join the OpenAI Forum on May 13 for a conversation with Thibault Sottiaux, Head of Codex, on how agentic tools can speed up tasks other than coding. This talk is for researchers, educators, and other knowledge workers who are interested in learning more about how AI agents can reduce friction in their work. The session will also include a live audience Q&A with the Thibault and an optional 1:1 networking session. 📅 Wednesday, May 13 🕒 1:45–2:55pm PDT 👉 Registration linked below

We're excited to announce that the OpenAI Academy and TBPN are teaming up to hold a event on how AI is actually being used inside fast-moving newsrooms. It's the Academy's first-ever collaboration with the show. In this upcoming OpenAI Academy for News session, OpenAI’s Evan Hirsch joins TBPN co-host and head of engineering Tyler Cosgrove for a behind-the-scenes look at how AI helps support three hours of live daily programming: from spotting angles and shaping coverage to building repeatable editorial systems. Attendees will leave with practical takeaways any newsroom or editorial team can use immediately; no technical background is required. And we’ll leave plenty of time for questions. https://lnkd.in/gphntAsC