Claude Mythos 5.2: How Access Works, What Changed, and Who Should Ask

Claude Mythos 5.2 is the version of Anthropic’s newest model with fewer cyber and biology safeguards, and only vetted organizations can use it. It launched on November 9, 2026, alongside Claude Fable 5.2. Anthropic says the two are the same underlying model. Fable 5.2 is the one most paying customers get.

We covered the launch in Claude Fable and Mythos 5.2 and the public model in our Claude Fable 5.2 deep dive. This post covers Mythos only: what it is, how the gate works, what Anthropic says changed since Mythos 5 and 5.1, what it costs, and who among our readers should ask for it.

The short version: most agencies, nonprofits and campus IT teams will not need Claude Mythos 5.2. A security team doing defensive work can apply through the Cyber Verification Program. An organization on Claude Enterprise can already get Mythos-powered code scans through Claude Security. Some of Anthropic’s help pages still describe Mythos 5.1, so read them as of today and expect updates.

What Claude Mythos 5.2 is

According to Anthropic’s launch announcement, "Claude Fable and Claude Mythos are the same underlying model." Fable 5.2 adds safeguards for high-risk cyber and biology work. Mythos 5.2 is "available to vetted organizations that work in these dual-use areas."

Anthropic calls it "our most capable model for cybersecurity engineering." The company’s system card, dated November 9, says Mythos 5.2 "has more permissive safeguards" in biology and cybersecurity. It does not list each safeguard that is relaxed.

The gap shows up in the requests Fable 5.2 refuses. Per the announcement, Fable 5.2 blocks high-risk cyber work such as penetration testing, exploit generation and binary vulnerability work. Those requests fall back to Claude Opus 4.8. Most research biology falls back to Claude Opus 5. A verified Mythos user can do that work on the newer model.

The specs of Claude Mythos 5.2 match Fable 5.2. Anthropic’s What’s new in Claude Fable 5.2 page lists a 1M-token context window, 128k max output and always-on adaptive thinking for both. The API model ID is claude-mythos-5-2.

How Claude Mythos 5.2 access works

The release notes say Mythos 5.2 is limited to "organizations verified through Anthropic’s verification programs, such as the Cyber Verification Program." The announcement names three routes in.

The third route needs care. Anthropic’s Claude Security page says the product scans your codebase, checks findings and recommends patches. It is in public beta for Claude Enterprise. Mythos powers the scans, but the scans run only in the Claude Security app.

When Mythos 5 came to Claude Security in August, Anthropic’s August 21 post said "The Mythos scan itself does not extend Mythos access to other surfaces." Nothing we read says that changed with 5.2.

The Life Sciences Verification Program form is a register-interest page for an organization’s admin. As of November 10, it still describes "expanded access to Mythos 5.1." It also says completing the form does not guarantee access.

The Cyber Verification Program, step by step

Anthropic’s Cyber Verification Program article explains the main route. As of November 10, it still names Claude Mythos 5.1, plus "new models moving forward." Here is what it says today.

• Apply once per organization through the Verification Portal. Describe your security work and attest to the controls for the tier you want.

• Three tiers: Defense, Red Team and Specialized. Red Team and Specialized are for organizations only. Each Specialized applicant is reviewed in depth with the US government.

• Individuals such as independent researchers and bug bounty hunters can apply for Defense Access only, on a paid plan.
  Once you apply, the process runs like this.

• Timing: Anthropic aims to send a decision or a request for more information within seven business days.

• Approval does not switch access on. An admin assigns it to a new API workspace or, on Enterprise, a custom role. On Pro, Max and Team, it attaches at the organization level.

• Cloud access to Mythos trails approval by about five business days.

• Security rules: Defense Access requires some form of MFA now, and API keys expire every seven days. By December 15, 2026, Defense members must use phishing-resistant MFA and stop using API keys.

The article says the program "is built for defense." Organizations whose main business falls outside that scope may not qualify. It names no country list.

Where Claude Mythos 5.2 runs

The What’s new page lists four places to run Mythos 5.2. It does not list Claude Platform on AWS, which does carry Fable 5.2.

The Bedrock limit has a reason. The cyber program article says Bedrock "does not yet support human review of automated safety flags, which CVP requires by default."

One page disagrees with another here. The cyber program article lists Claude Platform on AWS as a program surface, but the What’s new page does not list Mythos 5.2 there. Ask your account team before you plan around it.

What changed from Mythos 5 and Mythos 5.1

Mythos has moved fast this year. Mythos 5 launched on June 9 for a small partner group. It was suspended on June 12 under a US export-control order and returned in stages, as we covered in the June suspension post and the July restore post. Mythos 5.1 arrived on September 1 through the verification programs, per our Claude Fable 5.1 post.

Here is what Anthropic’s announcement reports for Claude Mythos 5.2. These are Anthropic’s own tests.

• Vulnerability finding: On its internal CVE Rediscovery test of 1,829 published CVEs, Mythos 5.2 found 10.2% more vulnerabilities than Opus 5.5 and 14.4% more than Mythos 5.1.
• Scenario test: On CyScenarioBench, Mythos 5.2 scored 76.3%, against 61.7% for Mythos 5.1 and 67.6% for Opus 5.5 without safeguards.
• Fewer overstated findings: One tester found earlier models overstated about 1 in 20 findings. Mythos 5.2 overstated none.
• Sandbox tests: Mythos 5.2 used credentials or files placed outside its sandbox in 3.1% of adversarial scenarios. The figures were 39% for Mythos 5 and 21% for Mythos 5.1.

The system card adds some limits. It says Mythos 5.2 "still falls within the lower category of capabilities" in Anthropic’s Frontier Compliance Framework. It refused malicious computer use tasks at rates among the highest of Anthropic’s recent models. In Claude Code, it refused malicious requests "somewhat less often than Claude Mythos 5.1."

Data handling has not changed. The release notes say both 5.2 models require 30-day retention and are not available under zero data retention unless Anthropic expressly authorizes it. The announcement says eligible customers can keep using both with zero data retention ahead of Enterprise Frontier Safeguards.

Claude Mythos 5.2 pricing

Anthropic’s pricing page gives Mythos 5.2 the same rates as Fable 5.2. Here is how it compares with earlier Mythos models, read on November 10.

Anthropic’s announcement says Mythos fast mode is "available only to organizations who are part of our Verification Programs." Claude Security scans bill as standard token usage under your plan, per the August post.

Who in our audience should ask for access

Start with the work you do. If your requests are not being blocked, Fable 5.2 or Opus 5.5 will do. Anthropic itself recommends Opus 5.5 as the starting point for most coding and knowledge work. Our Claude Opus 5.5 post covers it.

• Security teams and MSSPs that run penetration tests, triage malware or validate exploits for clients. This is the case the cyber program was built for.
• Agencies with a security practice. If you audit client sites, you could apply as an organization. You would also need to meet the MFA and API key rules.
• Higher-ed security offices and research computing. Campus security teams fit the cyber program. Biomedical labs may fit the Life Sciences program.
• Organizations already on Claude Enterprise. Turn on Claude Security first. It needs no cyber program membership. Our Claude Enterprise security controls post covers the admin side.

Most nonprofits and associations should not apply. A small team that patches a CMS and runs an office network does not need a model with fewer cyber blocks.

OpenAI runs a similar gate. Its GPT-6 Cyber model is limited to approved Daybreak Red customers.

What to do while you wait

1. Check whether you are blocked today. Run your security tasks on Fable 5.2. If they fall back to Opus 4.8 often, you have a case to apply.
2. Fix your sign-in setup first. Phishing-resistant MFA and keyless API access are coming for Defense members. Workload Identity Federation is Anthropic’s recommendation.
3. Apply once, as an organization. Pick one owner, and do not have staff apply separately.
4. Plan admin steps. Create a new API workspace for the program. Allow about five more business days for cloud access.
5. Settle retention. Confirm 30-day retention is acceptable for the code and logs you send.
6. Keep patching. Anthropic and its testers say Mythos 5.2 finds more vulnerabilities than earlier models, and those findings turn into patches you need to apply. Our guide on what a CVE is helps you read the advisories.

Frequently Asked Questions