OpenAI

We’re partnering with Amazon to accelerate AI innovation for enterprises, startups, and end consumers around the world. https://lnkd.in/g7y85Y-9

AI demand is growing rapidly as more people and businesses use AI to learn, build, and get work done every day. Today we’re announcing $110B in new investment, with support from SoftBank Group Corp., NVIDIA, and Amazon, to help scale the infrastructure needed to meet that demand and make AI more accessible, reliable, and useful for everyone. Hundreds of millions of people are using ChatGPT each week, and as usage grows, continued investment in compute, distribution, and partnerships is essential to bringing frontier AI to more communities worldwide. https://lnkd.in/gZ65Cjrh

We’re partnering with the Council of Fashion Designers of America (CFDA) to launch The Innovation Hub, a new two-year initiative designed to help American designers and brands explore the creative and business potential of AI. The program will pair promising CFDA members with leading AI builders to co-create tools that address real-world challenges facing fashion today in ways that are practical, responsible, and scalable. Our goal is to show how emerging technologies can meaningfully support creativity, strengthen business longevity, and drive positive change across the industry, while keeping designers firmly in the driver’s seat. https://lnkd.in/eZkxWB4e

🦞OpenClaw creator Peter Steinberger shares his best advice for developers today. On Builders Unscripted Ep. 1, he unpacks the journey of building OpenClaw, the power of open source, and how he uses Codex. See the full episode: https://lnkd.in/gkh_MEvJ

Stargate is the umbrella brand for our compute strategy. It’s about mobilizing the full ecosystem to deliver a step-change in global AI compute — and over the past few quarters that vision has become reality. Since announcing Stargate in January 2025, OpenAI has expanded from a mostly single-provider model to a broad partner network across clouds, silicon, and infrastructure: SoftBank, NVIDIA, AMD, Broadcom, Oracle, Microsoft, AWS, CoreWeave, and Cerebras. Demand is compounding, and no single provider, chip, or cloud can carry the load. A diversified compute model is the only way to scale at the pace AI requires. We exited 2025 with ~2 GW of available compute and a model designed to scale: long-term capacity agreements, purpose-built colo deployments, and deep collaboration on next-generation datacenter design. Compute leadership is the foundation of research and product velocity — and Stargate is how we build what comes next.

We’re committing $7.5M to The Alignment Project, a global fund for independent alignment research created by the UK AI Security Institute. As AI capabilities continue to advance, building a robust, diverse, and independent alignment ecosystem is essential. We look forward to continuing to collaborate with the broader research community as the field advances. https://lnkd.in/gd-Vkg_g

We’re introducing EVMbench, a new benchmark that measures how well AI agents can detect, exploit, and patch high-severity smart contract vulnerabilities. EVMbench measures three core capabilities: - Detect vulnerabilities in real-world contract code - Exploit them in realistic attack scenarios - Patch them safely, with fixes that hold up under testing EVMbench is intended both as a measurement tool and as a call to action. As agents improve, it becomes increasingly important for developers and security researchers to incorporate AI-assisted auditing into their workflows. We release EVMbench’s tasks, tooling, and evaluation framework to support continued research on measuring and managing emerging AI cyber capabilities. https://lnkd.in/gySryDDb

We’re building on our existing security protections with Lockdown Mode, a new advanced setting designed to help organizations mitigate prompt injection based data exfiltration attacks for employees who are especially at risk of being targeted. In these attacks, a third party attempts to mislead a conversational AI system into following malicious instructions to reveal sensitive information. Lockdown Mode in ChatGPT and Codex disables many network-enabled tools and some capabilities to provide stronger, more predictable security protections against prompt injection based data exfiltration. It’s intended for a small set of high-risk users, such as executives and security teams. We’re continuing to strengthen our safety and security safeguards to address novel, emerging, and evolving risks. This update—along with recent mitigations for URL-based data exfiltration—is a concrete step toward stronger protections against prompt injection. We’ll keep improving protections as AI systems grow more powerful. https://lnkd.in/e38RSQX8

GPT-5.2 derived a new result in theoretical physics. We’re releasing the result in a preprint with researchers from Institute for Advanced Study, Vanderbilt University, University of Cambridge, and Harvard University. It shows that a gluon interaction many physicists expected would not occur can arise under specific conditions. GPT-5.2 proposed a formula for a gluon amplitude later proved by an internal OpenAI model and verified by the authors. With the help of GPT-5.2, these amplitudes have already been extended from gluons to gravitons, and other generalizations are also on their way. These AI-assisted results, and many others, will be reported on elsewhere. We appreciate the community’s feedback, and we’re grateful to our collaborators for their partnership. https://lnkd.in/eCFErbn4