OpenAI delays GPT‑5.6 rollout while expanding Codex super‑app

OpenAI announced on Friday that its next‑generation model, GPT‑5.6, will not be publicly released until the White House approves a limited customer list, even as the company rolls out a major Codex desktop update that lets the AI agent control any app on a user’s computer. The delay underscores growing regulatory friction, while the Codex upgrade signals OpenAI’s push toward a “super‑app” that does work for users across personal and enterprise tasks.

Government‑mandated hold on GPT‑5.6

OpenAI confirmed that the Trump administration’s request to stagger the release of GPT‑5.6 will keep the model in the hands of a “small set of pre‑approved customers” for the coming weeks, after which the company hopes to broaden access [1]. The company’s blog stresses the step is “short‑term” and intended to avoid a permanent licensing regime, but it also highlights the lack of an established voluntary framework for model review [1]. GPT‑5.6 Sol is billed as the most capable version yet on cybersecurity, biology and agentic benchmarks, yet those capabilities remain unavailable to the broader market while the approval process unfolds.

Codex transforms into a cross‑app AI assistant

In parallel, OpenAI unveiled a “massive” update to its Codex desktop environment, now able to interact with any Mac or Windows application—clicking, launching and typing on behalf of the user—while the user continues working in the foreground [3]. The update also adds a built‑in web browser for previewing front‑end work and integrates the gpt‑image‑1.5 model for on‑the‑fly image generation. OpenAI reports that Codex has reached 3 million weekly developers, a jump from the previously cited “more than five million weekly active users” for its broader ChatGPT ecosystem [2][3]. This expansion aligns with OpenAI’s reported ambition to turn ChatGPT into a “super app” that handles scheduling, travel, coding and content creation across devices [2].

Competitive context

OpenAI’s Codex upgrade arrives as rivals such as Anthropic are rolling out similar desktop agents (Claude Cowork) that lack simultaneous background interaction across all apps [3]. By enabling Codex to operate across an entire computer ecosystem, OpenAI positions its agent as a more versatile tool for developers and enterprises, potentially widening its moat against competing AI‑assisted coding platforms. Meanwhile, the regulatory pause on GPT‑5.6 may give competitors a window to capture market share for advanced language models, especially as the White House’s executive order seeks to curb “cybersecurity concerns” of frontier AI [1].

What to watch

• Model rollout – OpenAI’s next public announcement on expanding the GPT‑5.6 customer list, expected in the coming weeks.
• Enterprise adoption – Uptake of the new Codex desktop features among enterprise developers, tracked via weekly active user metrics.
• Regulatory response – Any formalization of the “voluntary process” for AI model review by the Trump administration, which could set a precedent for future releases.

The juxtaposition of a government‑mandated delay for its flagship language model and a bold expansion of Codex’s cross‑app abilities highlights OpenAI’s dual strategy: navigating regulatory scrutiny while building the infrastructure for an AI‑driven super‑app that could reshape how work is done across both consumer and enterprise landscapes.