Anthropic Expands Mythos 5 Access to More Defenders, Unveils $35

What happened

Anthropic has moved Claude Security out of a narrow early-access phase and opened it to the broader Claude Enterprise customer base, while simultaneously announcing a thirty-five million dollar fund dedicated to open source security work. Both moves signal that the company is treating security tooling as a core product line rather than a side project, and that it intends to back that positioning with a significant capital commitment to the wider ecosystem.

This article walks through what Anthropic actually shipped, what changed under the hood, who is most directly affected, how to get hands-on access, and what open questions remain worth tracking. It is written for security engineers, platform teams running Claude Enterprise licenses, and open source maintainers who may be eligible for funding.

Anthropic announced two things at once. First, Claude Security, which had been in public beta restricted to Claude Enterprise customers, now performs codebase scans using Mythos 5 as its underlying model. Second, the company unveiled a thirty-five million dollar fund directed at open source security initiatives. The pairing is deliberate: by upgrading the scanning capability and funding the community that produces the software being scanned, Anthropic is positioning itself at both ends of the supply chain — the tooling that finds vulnerabilities and the ecosystem that fixes them.

How it works

The timing matters for enterprise buyers who have been evaluating Claude Security during its public beta period. A model upgrade mid-beta is not routine; it suggests Anthropic concluded Mythos 5 produces meaningfully better results on the code analysis tasks that Claude Security depends on, and that the performance difference was large enough to justify a mid-cycle change rather than waiting for a general availability release.

The concrete technical change is that Claude Security's codebase scanning now runs on Mythos 5 instead of whatever model version it previously used. For a security scanning tool, the model version directly determines how well the system reasons about code paths, identifies vulnerable patterns, and avoids false positives that generate analyst fatigue. Mythos 5 is Anthropic's current flagship model, and applying it to codebase analysis means the scanner can handle longer context windows and more complex multi-file reasoning than earlier versions allowed.

Why it matters

What has not changed is the product's distribution boundary. Claude Security remains in public beta, and that beta remains limited to Claude Enterprise customers. The upgrade to Mythos 5 is an improvement to what beta participants already have access to, not an expansion of who can join the beta. Teams already enrolled should see scan quality improve without any configuration change on their end.

The most immediate audience is security teams inside organizations that already hold Claude Enterprise subscriptions. If their team enrolled in the Claude Security public beta, they are now running scans on Mythos 5 and should evaluate whether detection quality or false positive rates have shifted. Teams that have a Claude Enterprise contract but have not yet opted into the Claude Security beta have a clear prompt to do so now that the model underneath it has been upgraded.

The thirty-five million dollar open source fund broadens the relevant audience considerably. Open source maintainers, security researchers, and projects that sit inside critical software supply chains are all plausible beneficiaries. Anthropic has not publicly specified eligibility criteria, grant sizes, or application mechanisms in this announcement, but any maintainer whose project affects downstream enterprise software has reason to watch how the fund's structure gets detailed.

Who is affected

Claude Security is available now in public beta to Claude Enterprise customers. The Mythos 5 model upgrade appears to be automatic for enrolled participants rather than requiring an opt-in toggle, meaning existing beta users are already on the new model. Organizations that want access should start with their Claude Enterprise account and look for the Claude Security beta enrollment path, which would have been introduced when the beta opened to Enterprise customers.

The thirty-five million dollar open source fund does not yet have a published application process based on available information. Interested maintainers or researchers should monitor Anthropic's official channels for fund structure announcements, including eligibility requirements, grant categories, and disbursement timelines. No dollar amounts per grant, no application deadline, and no specific focus areas within open source security have been confirmed publicly at this stage.

What to watch next

The most significant near-term signal will be when, and under what terms, Claude Security moves out of public beta. The Mythos 5 upgrade mid-beta suggests Anthropic is still actively iterating on the product, which could mean general availability is several months away or that the beta period is being used to stress-test quality at scale before a broader launch. Enterprise buyers evaluating security tooling vendors should track whether the beta opens to non-Enterprise tiers or remains a premium-tier exclusive.

On the fund side, the details that matter most are grant size distribution, whether the fund targets specific ecosystems or vulnerability classes, and whether recipients are required to publish findings. A thirty-five million dollar commitment is large enough to move real projects, but its actual impact depends entirely on how it is structured and governed. Watch for a formal program announcement that addresses those mechanics.