There's an artificial intelligence model that already exists. It's ready. OpenAI's engineers have tested it and its executives know exactly what it can do. And yet it remains out of everyone's reach. Not because of technical flaws or insufficient training. Someone else has to sign off first.

ChatGPT 5.6 is complete. The United States government still hasn't authorized its release.

This follows a sequence we already know. In the analysis of Claude and Anthropic, one company understood before anyone else the decisive weight of the regulator. It built institutional relationships while talking about responsible safety. That bet is now paying off while OpenAI waits for an external stamp of approval.

The Musk-Altman trial revealed what few said out loud. OpenAI had stopped being the nonprofit oriented toward the common good long before it admitted it. The dispute wasn't just about betraying a mission. It was about who controls the narrative of what's safe and what isn't.

What's happening with ChatGPT 5.6 is the natural consequence of those two stories. OpenAI has developed its most capable model to date. Instead of releasing it with a demo and a subscription plan, it has to wait. Some committee or process with no official name has to give the green light. This didn't happen with previous versions. The difference lies in how we perceive scale.

A language model stopped being a technological curiosity. Today it functions as infrastructure that drafts contracts, supports diagnoses, generates critical code, and advises on decisions. When a technology reaches that level, governments want a key to it. Not necessarily to use daily. Just to confirm it exists.

The trend lines up with what we've seen in previous regulations of messaging and platforms. Governments act when the issue truly matters to them. And oversight frequently seeks something beyond protecting citizens. It seeks to secure access to whatever happens inside the systems.

I keep exploring exactly what the government reviews before giving the green light. Possible military-use capabilities. Disinformation risks. Market concentration. Or simply the assurance that any sensitive finding will also be available to them. I don't have all the answers. This is more complicated than it looks.

OpenAI occupies an uncomfortable position it helped build. Years of rhetoric about responsibility and alignment with human values created the perfect argument for regulators to say oversight was welcome. It's hard to push back when your own messaging insisted that such oversight was necessary.

Anthropic navigates the same terrain with greater ease. Not because its models are less powerful, but because it integrated institutional relationships early on—something OpenAI was slow to prioritize. Some researchers have long pointed out that the gap between the two companies is more political than technical.

There's a clear parallel to what happened in the pharmaceutical industry when regulatory agencies began requiring formal approvals for products that already worked. The organizations that turned that process into a competitive advantage managed to concentrate the market. High approval costs function as a natural barrier. This confirms a historical constant: when the referee shows up, the rules change for everyone.

ChatGPT 5.6 waits. Anthropic holds onto its institutional edge. And the United States government has become the central player on a board that once pretended it didn't need referees.

What does this really mean for anyone who still believes in the promise of democratizing artificial intelligence?

Sources:

1. OpenAI — Official statements on model development and relationships with federal regulators (openai.com)

2. Anthropic — Documentation on safety policy and institutional relationships (anthropic.com)

3. The Verge / Wired — Coverage of the federal regulatory process for frontier AI models (2025-2026)

4. Previous analysis: OpenAI Trial: Musk vs. Altman for the Future of AI? — Yves Laurent, April 2026

5. Previous analysis: Claude and Anthropic: Separating Real Bias from Invented Agenda — Yves Laurent, March 2026