Both leading US labs are now confidentially filed, and both have shipped under government access limits
OpenAI filed confidentially for an IPO following Anthropic. Both have separately restricted frontier releases at government request — OpenAI limiting models to trusted partners, Anthropic holding Mythos 5 to approved organisations before a partial release. Public-market disclosure and classified capability holds are about to share a balance sheet.
The tension is not hypothetical. A public company must disclose material risks and material products. A lab operating under a voluntary pre-release government access arrangement has commitments pointing the other way, covering confidentiality, insider risk and nondisclosure.
The precedent already exists in both directions. The administration allowed Anthropic to release Mythos to some companies and agencies; OpenAI limited new models to trusted partners at government request, then later moved to a public GPT-5.6 release ending those limits. Each of those transitions is, for a listed company, a disclosable event with a classified input.
The covered frontier model designation sharpens it further. A classified determination that a model crosses a capability threshold is material information a company cannot fully explain to its shareholders.
TechCrunch — Following Anthropic, OpenAI files confidentially for IPO → · CNBC — Trump admin allows Anthropic to release Mythos AI model to some companies, government agencies → · CNBC — OpenAI limits new AI models to trusted partners at request of U.S. government →