AI development tools and technologies

Description

Appropriate tools, frameworks, and platforms for responsible AI development including fairness testing and explainability capabilities.

Implementation Guidance

Approved Toolchain

Define and maintain an approved list of AI development tools, frameworks, and platforms. Evaluation criteria should include: responsible AI capabilities (bias detection, explainability), security posture, license compliance, vendor stability, and integration with existing infrastructure.

Responsible AI Tooling

Require that approved toolchains include capabilities for: fairness testing (disparate impact analysis, demographic parity checks), explainability (SHAP, LIME, feature importance), robustness testing (adversarial inputs, edge cases), and privacy preservation (differential privacy, data anonymization).

Tool Governance

Establish a review process for adopting new AI tools. Prohibit use of unapproved tools for production AI systems. Conduct periodic security and compliance assessments of approved tools.

Evidence Requirements

• Approved AI tools and platforms register
• Tool evaluation criteria and assessment records
• Responsible AI capability requirements
• Tool adoption approval records