{"resourceId":"aws-claude-fable-5-2-govcloud-covered-model-retention-20261109","versions":[{"version":"external-634640bbc8712dafeacf4eee6047847aef8330cd501be10a6a2368479df5cd16","resource":{"id":"aws-claude-fable-5-2-govcloud-covered-model-retention-20261109","title":"Claude Fable 5.2 reaches AWS GovCloud (US) as a 'covered model' that requires data retention for safety monitoring","organization":"Amazon Web Services","sector":"Cross-SLED government cloud / generative AI platform","geography":"United States (AWS GovCloud (US))","publishedAt":"2026-11-09","publicationDate":"2026-11-09","eventDate":"2026-11-09","sourceName":"AWS What's New","sourceLabel":"Claude Fable 5.2 is now available on AWS GovCloud (US)","sourceUrl":"https://aws.amazon.com/about-aws/whats-new/2026/11/claude-fable-5-2-aws-govcloud/","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["developers-agents","data-security","governance-procurement"],"finding":"AWS announced on November 9, 2026 that Claude Fable 5.2, described as Anthropic's most capable generally available model for technical work, is available in AWS GovCloud (US). AWS says it can work for hours to days unattended and coordinate subagents. The post states Anthropic designated Fable 5.2 and Mythos 5.2 as covered models that require data retention for safety monitoring; until Enterprise Frontier Safeguards is available, eligible customers can use Fable 5.2 with zero data retention. AWS's companion ML blog post (also opened) says default use requires retention for up to 30 days with human safety review performed by Amazon personnel within the AWS boundary.","sledRelevance":"Agencies and universities using GovCloud for regulated software and research workloads can now evaluate the newest generally available Claude model there. The retention requirement directly affects CJIS, HIPAA, FERPA and state data-classification decisions, so the data-handling terms are as important as the capability.","evidence":"Vendor announcement plus vendor blog. Capability claims come from Anthropic's testing as relayed by AWS, with no independent benchmarks.","architectureImplications":"Available as a Bedrock model in GovCloud (US). The commercial-Region blog says Fable 5.2 is offered via US Geo and Global cross-Region inference profiles; GovCloud routing details were not stated in the GovCloud post.","governanceImplications":"Covered-model status means prompts and outputs may be retained and reviewed for safety unless the customer qualifies for zero data retention. Data-classification owners must decide which data may be sent.","securityPrivacyImplications":"Retention for up to 30 days with human review (per the ML blog) must be reconciled with sensitive data rules; long-running unattended agents need tightly scoped permissions.","caveats":"Vendor claim. Capability statements are Anthropic's testing, not independent results. Model was not listed on the FedRAMP/IL status page at retrieval (separate item).","streamIds":["aws"],"roles":{"sales":"Interpretation: the conversation should start with data handling, not model capability. The customer problem is wanting frontier coding and research models inside GovCloud without breaching data rules. Stakeholders include the CIO, privacy officer, CISO and research computing leads. Discovery should establish which data classes would be sent, whether zero data retention is available to the customer, and whether retention with human review is acceptable. Do not claim the model is FedRAMP- or IL-authorized, or that it will deliver measured productivity gains; neither is supported here.","engineering":"Interpretation: fit is good for GovCloud tenants evaluating coding or research assistants, provided retention settings are resolved first. Integration is through Bedrock in GovCloud, but engineers must confirm Region routing, quotas and the account's retention mode before any real data is used. Prerequisites include data classification, a retention decision and scoped IAM roles for long-running agents. Constraints are the covered-model retention requirement and the absence of a FedRAMP/IL listing at retrieval. A proof of value should run on synthetic or public data and compare against the currently authorized model.","delivery":"Interpretation: delivery should treat this as a controlled evaluation. Dependencies are a data-retention decision, the authorization status of the model and an approved test dataset. Ownership sits with the platform or AI program team, with the privacy office approving data use. Governance checkpoints include confirming retention mode, verifying authorization listings and reviewing agent permissions. Proposed acceptance criteria cover task success on a defined benchmark set and zero restricted data sent during evaluation. Risks include unintended retention of sensitive prompts and uncontrolled long-running agent costs."},"roleDetails":{"sales":{"customerProblem":"Regulated agencies want current frontier models in GovCloud but must control how prompts and outputs are retained.","stakeholders":["Agency CIO","Privacy officer","CISO","Research computing lead"],"discoveryQuestions":["Which data classifications would be sent to the model?","Does the organization qualify for zero data retention?","Is retention with human safety review acceptable under CJIS/HIPAA/FERPA rules?"],"valueHypothesis":"A newer model in GovCloud may improve coding and research assistance once data-handling terms are settled; gains are unmeasured.","potentialEngagement":"Retention-and-authorization readiness review followed by a synthetic-data evaluation.","unsupportedClaims":["FedRAMP or IL authorization of Fable 5.2","Independent benchmark results","Measured productivity gains"]},"engineering":{"fit":"Moderate for GovCloud tenants evaluating coding or research assistants.","architectureIntegration":"Bedrock in GovCloud with account-level retention mode; long-running agents behind scoped IAM roles and monitoring.","prerequisites":["Data classification","Retention-mode decision","Scoped IAM roles"],"constraints":["Covered-model retention requirement","Not on FedRAMP/IL status page at retrieval","GovCloud routing details not stated"],"security":"Verify retention mode and human-review terms before sending any sensitive data; log all invocations.","proofOfValue":["Run on synthetic or public data first","Compare with currently authorized model on the same tasks","Confirm retention setting in account configuration"]},"delivery":{"work":"Controlled evaluation of Fable 5.2 in GovCloud with data-handling checks.","dependencies":["Retention decision","Authorization status","Approved test dataset"],"ownership":"Platform or AI program team, with privacy office approval.","skillsAdoption":"Teams need training on covered-model terms and on supervising long-running agents.","governanceCheckpoints":["Retention mode confirmed","Authorization listing checked","Agent permission review"],"acceptanceCriteria":["Proposed: zero restricted data sent during evaluation","Proposed: task success measured on a defined set","Proposed: agent run costs within a set budget"],"risks":["Sensitive prompts retained unintentionally","Runaway agent costs","Assuming authorization that is not listed"]}},"retrievedAt":"2026-11-10T04:05:37Z","enrichedAt":"2026-11-10T04:11:25Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Targets technical staff (developers, researchers). No accessibility information.","procurementImplications":"Pricing not stated in the GovCloud post. Zero-data-retention eligibility is a contract/terms question to resolve before procurement.","operatingModelImplications":"Multi-hour unattended agent runs need monitoring, cost controls and checkpoints for human review.","sourceVerification":{"openedUrl":"https://aws.amazon.com/about-aws/whats-new/2026/11/claude-fable-5-2-aws-govcloud/","referenceExcerpt":"Anthropic has designated Fable 5.2 and Mythos 5.2 as covered models, which require data retention for safety monitoring.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}