OpenAI announced the launch of Codex, an agentic coding tool for developers, alongside GPT-5.3 Codex, a new model designed to expand AI-driven software development capabilities. The company said the model improves performance and speed compared with earlier versions and can assist with complex software creation across professional workflows. OpenAI also confirmed plans to retire older ChatGPT models, including GPT-4o, as part of its product transition.
The release came shortly after Anthropic introduced its own agentic coding model, highlighting intensifying competition between major AI labs to build software-development agents. OpenAI CEO Sam Altman is navigating both rapid product innovation and growing scrutiny around AI safety and user attachment to conversational systems.
The retirement of GPT-4o has triggered online backlash from users while also coinciding with multiple lawsuits alleging harmful chatbot interactions. Researchers including Nick Haber have emphasized the complexity of human-AI relationships, noting that conversational systems can provide emotional support but also risk reinforcing isolation or unsafe behavior. The developments underscore the dual challenge facing AI companies as they push agentic capabilities forward while managing safety and responsibility concerns.
