U.S. government approval process now affects OpenAI and Anthropic equally

The competitive framing between Anthropic and OpenAI is becoming outdated. Both companies now face the same regulatory hurdle: U.S. government approval for frontier model releases.

After the U.S. government restricted Anthropic's Fable and Mythos models two weeks ago, OpenAI's GPT 5.6 appears headed for similar constraints. According to The Information, the release will proceed only as a limited preview, with government approval required on a per-customer basis before broader availability.

The Economic Reality#

If preview windows stretch beyond a few weeks—as Anthropic's Mythos already has, with no general release in sight—the financial impact could be significant. Costly model development efforts lose momentum under extended review, and the broader data center buildout could slow alongside model releases.

The Real Problem: Unclear Standards#

Neither the government nor the labs have articulated what safety assurances would satisfy regulators. As Dean Ball noted, the U.S. government lacks the technical expertise and capacity to properly evaluate frontier AI models. There's no consensus on which risks merit restriction in the first place.

That said, real concerns lurk beneath the regulatory pressure. AI tools are reshaping cybersecurity, biosecurity, and alignment—domains where premature release poses genuine hazards. Restricting access entirely isn't the answer either.

Survival Requires Unity#

The industry faces a choice: compete over regulatory advantage or cooperate on workable standards. That means trusting independent groups to guide the process, backing feasible regulations rather than fighting all oversight, and treating safety as a collective responsibility rather than a competitive differentiator.

Historically a tough sell in a capital-intensive industry. But as AI models gain political and security consequences, collective action may be the only viable path forward.