Amazon Bedrock now offers Anthropic’s Claude Mythos 5.2 in AWS GovCloud (US), AWS says. The model is available only to eligible, allow-listed organizations, which must go through Anthropic’s Cyber Verification Program to get started. This is the first Anthropic model AWS describes as having powerful cybersecurity capabilities to be offered in GovCloud.
AWS describes Mythos 5.2 as Anthropic’s most advanced model to date for cybersecurity, software coding and complex reasoning. The company says it can identify sophisticated software vulnerabilities and demonstrate how they could be exploited. It can also examine large codebases and return actionable findings with less manual guidance than previous AI models. These are AWS’s descriptions; the announcement does not include test results or examples showing how well the model performs.
What AWS says it can do
AWS presents the model as a tool for defensive security work. Security teams could use its findings to locate and fix vulnerabilities in software, including critical software, before those weaknesses are exploited. The announcement does not say which software systems have been tested, how many vulnerabilities the model has found, or whether its findings have been independently checked. It also gives no comparison figures for the claimed reduction in manual guidance.
That distinction matters: identifying a possible weakness and demonstrating exploitability can help security teams investigate a risk, but the announcement does not establish that the model’s findings are always correct or that using it prevents attacks. AWS does not publish benchmark scores, pricing or other details about how customers will be charged for access.
Access is restricted
The release is not a general launch for every Amazon Bedrock customer. Access is limited to organizations approved through Anthropic’s Cyber Verification Program, and the model is offered in AWS GovCloud (US) Regions. AWS does not list the eligibility criteria or say how long approval takes. For organizations that are not allow-listed, the announcement provides no route to immediate access.
For approved security teams, the practical change is that they can use Mythos 5.2 through Bedrock in GovCloud rather than relying on a general-purpose coding assistant alone. The announcement does not describe any additional setup requirements or explain how the model fits into existing security workflows. Teams will need to assess its findings rather than treat them as confirmed vulnerabilities.
Why it matters
Approved security teams using AWS GovCloud can access a model AWS says can help inspect large codebases and flag exploitable vulnerabilities. Access is restricted, and the announcement does not establish how accurate its findings are or how much time it saves.
