Prediction market Polymarket gives only a 17% probability to Anthropic releasing Claude Mythos by 30 June, but puts the odds at 65% for a July launch, pointing to a mid-summer public debut for one of the most controversial AI models ever developed.
With Anthropic having filed confidentially for an IPO targeting an October listing at a valuation approaching $1 trillion, the timing is unlikely to be coincidental.
Mythos has had a turbulent journey into public consciousness.
The model's existence first emerged in late March 2026 not through an announcement but through a configuration error in Anthropic's content management system, which left nearly 3,000 unpublished internal assets, including draft blog posts describing the model, in a publicly accessible data store.
Fortune reviewed the material before Anthropic removed access. The company confirmed the model was real, describing it as a step change in AI capability with meaningful advances in reasoning, coding and cybersecurity, and its most capable model to date.
The nature of those cybersecurity capabilities immediately generated alarm.
Leaked documents suggested Anthropic internally believed Mythos posed unprecedented cybersecurity risks, and the company was privately warning government officials that the model made large-scale cyberattacks significantly more likely. Cybersecurity stocks fell on the news.
Matters were compounded in April when, on the very day Anthropic launched Claude Mythos Preview and announced Project Glasswing, its restricted partner programme granting a small group of major organisations early access for defensive security work, a group of unauthorised users in a private Discord server gained access to the model through a third-party vendor environment.
Anthropic said it found no evidence that the breach extended beyond that environment.
The model has since remained tightly restricted, available only to a handful of companies, including Microsoft, Apple and Google under the Project Glasswing framework.
But with a July public release now the market consensus, and an IPO roadshow to follow in the autumn, Anthropic has every incentive to demonstrate Mythos to the broadest possible audience before it asks public investors to value the company at ten times what it was worth eighteen months ago.
Whether a model described internally as an unprecedented cybersecurity risk can be safely released at scale remains the rather large question that Polymarket's odds do not attempt to answer.
