Dimension 01
Data handling & privacy
What data does the tool collect, where is it stored, who can access it, and does the vendor commit — contractually — not to use it for model training?
Methodology
Responsible AI Rubric
Every review on this site uses the same five-dimension framework. The goal is consistency: the same criteria, applied with the same rigor, whether the tool is free or expensive, well-known or new.
Dimension 02
Transparency
Can a user or reviewer understand how the model works, what it was trained on, and what its limitations are? Black-box claims without evidence score lower.
Dimension 03
Bias mitigation
Has the vendor published evidence of testing for harmful bias, and are there mechanisms for users to report and escalate problematic outputs?
Dimension 04
Human oversight
Does the tool keep a human in the loop for high-stakes decisions, and does it make clear when output should be reviewed rather than acted on directly?
Dimension 05
Vendor accountability
Are there clear terms of service, a published security posture, independent audits, and a responsive process for incidents and questions?
How scores map to verdicts
Each dimension is scored against publicly available evidence. A tool's total score out of 30 determines whether we recommend it, approve it with conditions, or reject it. Scores are not final certifications — they are a documented starting point for your own decision.