OpenAI AI Updates: July 16, 2026

1. OpenAI Introduces GPT-Red for Automated Red Teaming

OpenAI. OpenAI detailed GPT-Red, an automated red teaming system that uses self-play to have models generate and defend against adversarial attacks, aiming to strengthen robustness and safety without relying solely on human red teamers. The company frames the approach as a way to scale the discovery of model vulnerabilities and feed the resulting fixes back into training. It positions self-improvement as a core lever for making frontier systems more resilient. Source

2. OpenAI Argues State Laws Can Anchor National AI Safety Standards

OpenAI. OpenAI published a policy piece describing a “reverse federalism” framework in which state-level AI laws help establish national governance standards, with state and federal action reinforcing each other. The post lays out how the company believes AI safety regulation should be coordinated across US jurisdictions rather than left fragmented. It comes as several states move ahead with their own AI rules. Source

3. OpenAI Publishes Guidance on Managing AI Investments in the Agentic Era

OpenAI. OpenAI released guidance for enterprises on measuring productivity gains and optimizing AI spending as agents take on more of the workflow. The piece emphasizes scaling successful workflows rather than one-off deployments and offers a way to think about return on investment as AI agents move into production. It is aimed at leaders weighing where to expand AI budgets. Source

4. Codex CLI 0.144.5 Hardens Dangerous-Command Detection

OpenAI. OpenAI shipped Codex CLI 0.144.5, a patch release focused on security that improves dangerous-command detection, including additional forced rm forms, and provides clearer feedback when an operation is blocked. The change continues a steady cadence of maintenance releases for the Codex command-line tool. It reduces the risk of destructive commands running unchecked during automated coding sessions. Source

5. OpenAI Ships a $230 Light-Up Keyboard for Codex

OpenAI. OpenAI released a $230 dedicated keyboard for its Codex coding assistant, featuring illuminated keys mapped to Codex workflows, even as the company fights an Apple trade-secret lawsuit over its broader hardware ambitions. The peripheral is OpenAI’s first shipping hardware accessory and doubles as a marketing vehicle for Codex among developers. It lands amid reports of a more ambitious screenless AI device still in development. Source