Unreleased Mythos 1 briefly appeared inside Claude Code, outlets report

Anthropic’s unreleased Mythos 1 briefly appeared inside Claude Code and Claude Security over May 23 and May 24, according to reports from TestingCatalog and India Today. The sightings were short-lived, but both outlets said users saw model labels including “Mythos 1” and “claude-mythos-1-preview” before the references disappeared. Anthropic had said a day earlier that Mythos was unlikely to reach the general public soon and that wider release would require stronger safeguards. Fortune had previously reported on March 26 that Anthropic was testing a previously undisclosed model called Claude Mythos with early-access customers after draft materials were exposed in a public data cache. An Anthropic spokesperson told Fortune at the time that the model represented “a step change” in performance and was the company’s “most capable” system to date. (testingcatalog.com) ### Where did the new Mythos 1 references show up? TestingCatalog reported on May 23 that some users were temporarily able to see “Mythos 1” in the Claude interface and that source code included the line “Access to the Claude Mythos model in Claude Code and Claude Security.” The same report said Anthropic was also building out a Claude Security dashboard with vulnerability findings, seven-day and 30-day charts, and triage results for enterprise users. (tech.yahoo.com) India Today reported on May 24 that users on X said they saw both “Mythos 1” and “claude-mythos-1-preview” on Claude Code and Claude Security. The outlet said the references appeared only briefly and were later removed. ### Why did the sightings draw attention this weekend? Anthropic said on May 22 that Mythos-class models could reach general release only after the company had developed “far stronger safeguards,” according to language cited by TestingCatalog and India Today. (testingcatalog.com) India Today reported that Anthropic had limited access to selected partners and said public release was unlikely for now. (indiatoday.in) The timing mattered because the product traces appeared one day after that public caution. Neither report said Anthropic had announced a launch date, and both framed the sightings as evidence of testing rather than a confirmed rollout. ### What has Anthropic said about Mythos before this? Fortune reported in March that Anthropic acknowledged developing and testing Claude Mythos after unpublished materials became publicly accessible because of what the company called a “human error” in its content management system. (testingcatalog.com) Fortune said the leaked draft described Mythos as more capable than Anthropic’s previous releases and linked it to a new “Capybara” tier above Opus. Anthropic also said the model was being trialed by “early access customers,” according to Fortune’s report. That earlier disclosure established that Mythos existed and was already in limited testing, even though Anthropic had not formally launched it. ### Does this mean a public launch is imminent? TestingCatalog said the new strings and interface references pointed to preparation for Claude Code and Claude Security, but it also said there were “no indications yet” that the product would move beyond enterprise customers. (tech.yahoo.com) India Today similarly reported that there was “no clear indication” Mythos 1 would actually go to wider release and noted Anthropic had removed the traces. Anthropic’s own public position, as quoted by the outlets, remains that Mythos-class models would come later only once stronger safeguards are in place. For now, the clearest next marker is whether Anthropic leaves Mythos 1 references visible in Claude Code or Claude Security, or formally expands access beyond current Project Glasswing and early-access partners. (testingcatalog.com)