The Technology Advisory Panel (TAP) has published an AI Proportionality Assessment Aid, a framework designed to guide the proportionality assessment of AI models.

This aid offers a structured approach to ensure AI models are the appropriate tool for their intended tasks.

The TAP serves as an informational panel for the Investigatory Powers Commissioner’s Office and the Judicial Commissioners on the impact of changing technology and the development of techniques to use these tools under prescribed investigatory powers whilst minimising privacy interference.

The AI Proportionality Assessment Aid is a practical framework aimed at guiding the proportionality assessment of data-driven AI models.

It outlines key questions to consider throughout the AI model lifecycle: concept, development, operation, and exploitation of results.

This framework builds on the recent CETaS (Centre for Emerging Technology and Security, The Alan Turing Institute) structured framework for assessing privacy intrusion, incorporating additional considerations and organising them according to the lifecycle stages.

It serves as a valuable resource for those involved in commissioning, designing, developing, testing and managing the governance of models.

The framework includes four stages to consider when assessing different AI tools.

Recommended reading

• PwC Interview | How Agentic AI is Reshaping Business
• The State of AI in 2025 | What You Need To Know
• Agentic AI on the Rise Despite Workers’ Concerns

Concept stage: This stage evaluates the necessity and appropriateness of AI approaches, considering biases, constraints and policy for re-training/testing.

Development stage: This focuses on data sensitivities, human inspection, data volumes, biases, and integrity ensuring robust data management practices.

Operation stage: This assesses the appropriateness of model use, data provenance, result delivery, and human inspection requirements.

Exploitation stage: This stage examines access to results, training for appropriate use, model chains, and reporting processes following adverse outcomes.

“I am delighted to publish the Technology Advisory Panel’s AI Proportionality Assessment Aid,” Dame Muffy Calder, chair of the TAP, said.

“This document provides a practical and comprehensive framework for assessing the proportionality of AI models, ensuring they are the appropriate tool for the task at hand.

“It is designed to guide stakeholders through the AI model lifecycle, from concept, to model development, application, and exploitation of results.

“The focus is on data-driven models. We believe this aid will be an invaluable resource for those involved in the development, governance, and application of AI technologies”.