OpenAI AI Updates: June 24, 2026

1. OpenAI Calls for Shared Standards on Advanced AI

OpenAI published a piece on helping build shared standards for advanced AI, outlining its position on common frameworks for developing and governing frontier systems. The post is part of OpenAI’s broader engagement on AI policy and coordination across the industry. It signals continued effort to align practices around safety and accountability as model capabilities advance. Source

2. GPT-5 Helps Immunologist Resolve a Three-Year Research Mystery

OpenAI shared a case study describing how immunologist Derya Unutmaz used GPT-5 to solve a research question that had remained open for three years. The account details how the model assisted in working through the problem in his immunology research. It is presented as an example of GPT-5 contributing to scientific investigation. Source

3. Omio Builds Conversational Travel Experience on OpenAI Models

OpenAI highlighted how travel booking platform Omio is using its models to build conversational travel features. The piece describes Omio’s approach to letting users plan and book trips through natural language interactions. It serves as a customer example of OpenAI technology applied to consumer travel. Source

4. ChatGPT for iOS Adds Personality Settings and Goal Editing

OpenAI shipped an update to ChatGPT for iOS that adds personality settings, goal editing, and improved chat visibility. The release also includes fixes for Face ID and dark mode indicators. The changes expand customization and reliability for mobile users. Source

5. Codex CLI 0.142.0 Adds Usage Credits and Multi-Agent Controls

OpenAI released Codex CLI version 0.142.0 with usage credit redemption, reorganized plugin management across curated, workspace, and shared sections, and configurable token budgets for agent operations. The release also introduces multi-agent delegation controls at the thread and turn levels. The update gives developers finer control over how Codex agents spend tokens and coordinate work. Source

6. OpenAI Says GPT-5.5-Cyber Beats Anthropic’s Mythos on Cyber Benchmark

OpenAI announced GPT-5.5-Cyber, a model variant specialized for cybersecurity tasks that the company says outperforms Anthropic’s Mythos on a cybersecurity benchmark. The release reflects intensifying competition over security-focused model variants and their dual-use offensive and defensive capabilities. The benchmark claim is OpenAI’s own and has not been independently verified. Source

7. OpenAI Launches Initiative to Find and Patch Open Source Bugs

OpenAI launched an initiative that uses its models to discover and patch bugs in open source software. The effort extends the company’s push into AI-assisted security and code maintenance for widely used dependencies, where a single vulnerability can propagate across the ecosystem. It positions OpenAI alongside other labs treating open source security as a proving ground for autonomous code agents. Source