Fortune reported on March 26, 2026 that Anthropic was developing Claude Mythos and had begun testing it with early-access customers after a data leak revealed the model’s existence. In other words, the reported path was early access first, then preview—not an announced public launch.
The confusion comes from the way AI model launches are often described. “Preview,” “released,” “announced,” and “available” can refer to very different access levels. For Claude Mythos, the available sources consistently point to restricted access rather than a normal public release.
Security is the central caveat in the reporting. Fortune said leaked draft material described Claude Mythos as posing “unprecedented cybersecurity risks,” and Forbes and Business Insider also framed broader release around safety or security concerns. That does not prove Mythos will never become public, but it does explain why the available reporting does not show a standard launch date.
A genuine public release would need clearer signals than “preview” or “early access.” Useful indicators would include:
Until those signals appear, “available” should be read carefully. In the current reporting, Mythos appears available only in limited or preview contexts, not as a broadly accessible Claude model.
Claude Mythos has no confirmed public release date in the available reporting. The best-supported timeline is that Anthropic was testing Mythos with early-access customers by March 26, 2026, and debuted a preview on April 7, 2026, while broader availability remained unconfirmed.