Cada elemento de los frameworks, mapeado al o los controles de seguridad de AWS que lo atienden: OWASP LLM Top 10 (2026) para riesgos técnicos, NIST AI RMF para gobernanza del riesgo, las mejores prácticas de seguridad de la AWS Well-Architected Generative AI Lens y los checks de AWS Security Hub AI para validación automatizada en runtime. Úsalo para ver la cobertura de un vistazo y rastrear cada control hasta el framework.
ⓘ Esto mapea cada elemento de los frameworks (riesgos del OWASP LLM Top 10, funciones/categorías del NIST AI RMF, mejores prácticas de la AWS Well-Architected Generative AI Lens y checks de AWS Security Hub AI) a uno o más controles de seguridad de AWS de este modelo. Muestra cobertura y trazabilidad de controles; no es una declaración de cumplimiento ni certificación.
AWS Well-Architected Generative AI Lens (Security) 10/10 elementos cubiertos
OWASP GenAI/LLM Top 10 (2026) 10/10 elementos cubiertos
OWASP GenAI Top 10 for Agentic Applications (2025) 10/10 elementos cubiertos Elemento del framework Controles de seguridad AWS ASI01 Agent Goal Hijack c4.5 c5.6 c5.8 ASI02 Tool Misuse and Exploitation c4.1 c4.4 c4.5 c5.8 ASI03 Identity and Privilege Abuse c4.1 c4.2 c5.1 c5.4 ASI04 Agentic Supply Chain c5.10 ASI05 Unexpected Code Execution c4.7 ASI06 Memory and Context Poisoning c4.3 c5.2 c5.3 ASI07 Insecure Inter-Agent Communication c4.6 c5.7 ASI08 Cascading Failures c4.8 c5.9 ASI09 Human-Agent Trust Exploitation c5.5 ASI10 Rogue Agents c5.10
NIST AI Risk Management Framework (AI 100-1) 15/19 elementos cubiertos El NIST AI RMF se mapea a nivel de función y categoría (Govern, Map, Measure, Manage). No toda categoría aplica a un modelo de seguridad técnico, así que una cobertura parcial aquí es esperada y honesta.
Elemento del framework Controles de seguridad AWS GOVERN.1 Policies, processes, procedures and accountability structures c1.1 c1.11 c2.8 c4.1 c5.1 c6.8 GOVERN.2 Accountability structures, roles, responsibilities and training Sin control mapeado aún GOVERN.3 Workforce diversity, equity, inclusion and interdisciplinary teams Sin control mapeado aún GOVERN.4 Organizational culture that prioritizes risk awareness c1.11 GOVERN.5 Processes for stakeholder engagement and feedback c5.5 GOVERN.6 Policies for third-party software, data and supply chain risk c2.6 c2.8 MAP.1 Context is established and understood Sin control mapeado aún MAP.2 AI system categorization and capabilities are documented c3.8 MAP.3 AI benefits and costs are understood Sin control mapeado aún MAP.4 Risks and impacts are mapped c1.8 c2.5 c2.6 c3.4 c3.5 c5.8 MAP.5 Impacts to individuals, groups and society are characterized c3.5 MEASURE.1 Appropriate methods and metrics are identified and applied c1.12 c2.5 c5.10 MEASURE.2 AI systems are evaluated for trustworthy characteristics c1.3 c1.9 c1.12 c2.9 c3.9 c4.5 c5.6 c5.10 c6.3 MEASURE.3 Mechanisms for tracking AI risks are in place c1.4 c1.10 c2.7 c2.9 c3.8 c4.8 c5.9 c6.7 c6.8 MEASURE.4 Feedback about efficacy of measurement is gathered c5.10 MANAGE.1 AI risks are prioritized, responded to and managed c1.1 c1.2 c1.3 c1.5 c1.6 c1.7 c1.8 c1.9 c1.11 c2.1 c2.2 c2.4 c3.1 c3.2 c3.3 c3.4 c3.6 c3.7 c3.9 c4.1 c4.2 c4.3 c4.4 c4.5 c4.6 c4.7 c5.1 c5.2 c5.3 c5.4 c5.6 c5.7 c5.8 c6.1 c6.2 c6.3 c6.4 c6.5 c6.6 MANAGE.2 Strategies to maximize benefits and minimize impacts are planned c5.5 MANAGE.3 Third-party risks and benefits are managed c2.6 c2.8 MANAGE.4 Risk treatments, monitoring and response are documented and applied c1.4 c1.6 c1.10 c2.3 c2.7 c2.9 c4.8 c5.5 c5.9 c6.7 c6.8
AWS Security Hub — AI Security Best Practices 29/31 elementos cubiertos Estos son los checks automatizados de Security Hub CSPM que validan en runtime los controles correspondientes de este modelo. Dos checks (SageMaker.4, SageMaker.8) son de resiliencia y ciclo de vida de plataforma, no de postura de seguridad, así que quedan intencionalmente sin mapear.
Elemento del framework Controles de seguridad AWS Bedrock.1 Amazon Bedrock data sources should be encrypted with customer managed AWS KMS keys c1.2 c3.1 BedrockAgentCore.1 Bedrock AgentCore runtimes should be configured with VPC network mode c4.6 c5.7 BedrockAgentCore.2 Bedrock AgentCore Gateways should require authorization for inbound requests c4.4 c5.4 BedrockAgentCore.3 Bedrock AgentCore Memory should be encrypted with customer managed AWS KMS keys c5.2 BedrockAgentCore.4 Bedrock AgentCore Gateway should be encrypted with customer managed AWS KMS keys c4.3 BedrockAgentCore.5 Bedrock AgentCore custom browsers should not use public network mode c4.6 BedrockAgentCore.6 Bedrock AgentCore custom browsers should have session recording enabled c4.8 c5.9 BedrockAgentCore.7 Bedrock AgentCore custom code interpreters should use a private network configuration c4.7 SageMaker.1 Amazon SageMaker notebook instances should not have direct internet access c2.4 SageMaker.2 SageMaker notebook instances should be launched in a custom VPC c2.4 SageMaker.3 Users should not have root access to SageMaker notebook instances c2.2 SageMaker.4 SageMaker endpoint production variants should have an initial instance count greater than 1 Sin control mapeado aún SageMaker.5 SageMaker models should have network isolation enabled c2.4 SageMaker.8 SageMaker notebook instances should run on supported platforms Sin control mapeado aún SageMaker.9 SageMaker data quality job definitions should have inter-container traffic encryption enabled c2.4 SageMaker.10 SageMaker model explainability job definitions should have inter-container traffic encryption enabled c2.4 SageMaker.11 SageMaker data quality job definitions should have network isolation enabled c2.4 SageMaker.12 SageMaker model bias job definitions should have network isolation enabled c2.4 SageMaker.13 SageMaker model quality job definitions should have inter-container traffic encryption enabled c2.4 c2.9 SageMaker.14 SageMaker monitoring schedules should have network isolation enabled c2.4 SageMaker.15 SageMaker model bias job definitions should have inter-container traffic encryption enabled c2.4 c2.9 SageMaker.16 SageMaker models should use private registry in VPC for primary containers c2.4 SageMaker.17 SageMaker feature group offline stores should be encrypted with AWS KMS keys c2.1 SageMaker.18 SageMaker feature group online stores with standard storage should be encrypted with AWS KMS keys c2.1 SageMaker.19 SageMaker models should use private registry in VPC for multi-container inference pipelines c2.4 SageMaker.20 SageMaker model explainability job definitions should have network isolation enabled c2.4 SageMaker.21 SageMaker notebook instances should be encrypted with customer managed AWS KMS keys c2.1 SageMaker.22 SageMaker monitoring schedules should have inter-container traffic encryption enabled c2.4 SageMaker.23 SageMaker inference experiments should have instance storage volume encrypted with customer managed AWS KMS keys c2.1 SageMaker.24 SageMaker inference experiments should have data storage encrypted with customer managed AWS KMS keys c2.1 SageMaker.25 SageMaker model quality job definitions should have network isolation enabled c2.4 c2.9
Dimensiones de seguridad agéntica 6/6 elementos cubiertos 