Users briefly spotted a model labeled “Mythos 1” inside Anthropic’s Claude Code interface on May 24, one day after the company publicly stated that Mythos would remain restricted to roughly 40 partner organizations. Source code strings referencing “Access to the Claude Mythos model in Claude Code and Claude Security” also appeared, according to TestingCatalog, which first reported the sighting.

What Appeared

The model surfaced under the identifier “claude-mythos-1-preview” in the Claude Code model selector, according to India Today. The sighting was temporary. Anthropic removed traces of the model from the interface shortly after users flagged it on X and Telegram channels.

Separately, new strings appeared in Claude Code’s source code referencing both Claude Code and Claude Security as Mythos integration targets. Claude Security is Anthropic’s enterprise vulnerability triage product, which TestingCatalog reported is getting a dashboard refresh with seven-day and thirty-day historical charts for discovered vulnerabilities.

The Timing Contradiction

On May 22, Anthropic published a Project Glasswing update announcing that Mythos had helped discover more than 10,000 high- or critical-severity vulnerabilities in essential software since the program launched. The same update included this statement: “In the near future, once we’ve developed the far stronger safeguards we need, we look forward to making Mythos-class models available through a general release.”

That language represented a shift from Anthropic’s earlier framing, which positioned Mythos as permanently restricted. But the company still emphasized that the model was not ready for public access. The UI leak less than 48 hours later suggests that internal preparation for a broader rollout is further along than the public messaging indicates.

Prior Sightings

This is not the first time Mythos has appeared outside its intended restricted access. India Today noted that unauthorized users previously reported access to the model, and traces have surfaced on Google Cloud and AWS through vulnerability discovery programs. Claude Code’s source code was also leaked online earlier this year, exposing internal model references.

Claude Opus 4.8 Also Reportedly in Testing

In parallel, TestingCatalog reported that Claude Opus 4.8 is rumored to be in internal evaluation with select Anthropic partners, following the April release of Opus 4.7. A launch in the coming weeks would fit the cadence Anthropic has established for its Opus model line.

The Product Integration Signal

The combination of Mythos integration targets, specifically Claude Code for developers and Claude Security for enterprise buyers, suggests Anthropic is building two distinct distribution channels for its most capable model. Claude Code gives developers agentic coding capabilities powered by Mythos-level reasoning. Claude Security gives enterprise security teams autonomous vulnerability discovery and triage. Both channels align with Anthropic’s stated goal of making frontier capabilities available where they deliver the highest defensive value.

Anthropic has not commented on the UI sighting or confirmed a timeline for Mythos 1 availability.