AI Evidence Analyzer uses a variety of AI models, currently GPT-4o mini, to evaluate user-submitted evidence and Accountability Mechanisms for controls in the following conditions:
- Controls with questions have AM defined in the system
- The controls already answered and submitted, or under “Met by Evidence” / “Partially Met”
The evidence includes:
- Description (notes) in the Answer
- URLs and documents attached with the answer
- Description in the Accountability Mechanism selected in the response
- URLs and documents in the Accountability Mechanism selected in the response
AI evaluates the evidence with guidelines, including control guidelines, control descriptions, and legal citations from the privacy knowledge team. AI determines which guideline sections are addressed by the provided evidence, generates a coverage score, and highlights both fulfilled and unfulfilled areas. It also offers targeted recommendations for improvement, ensuring comprehensive compliance and helping to address any gaps effectively.
AI evaluation generates the following qualitative scores:
- Excellent: All major parts of the guidelines are covered with no significant gaps.
- Good: Most parts of the guidelines are covered, with minor gaps.
- Fair: Some parts of the guidelines are covered, but there are notable gaps.
- Poor: Many parts of the guidelines are not covered, with significant gaps.
- Unaddressed: The guidelines are not adequately covered, with many unresolved issues.
- Evidence Required - The initial state before the user attaches any evidence items.
- Evidence Not Required - These controls require no evidence; this classification still shows even if an evidence is attached.
- No URL/Attachments - These controls have accountability mechanisms attached but no URL/Attachments.
On the All-Questions View
The AI Evidence Analyzer is displayed in each control on the All-Questions View.
You can click on the result graphic to display more information.
Note that this result is not the final effectiveness score and just an AI recommendation. You can also provide feedback on the result by giving it a thumb up/down.
On the Review page
On the Review page, an AI Evidence Score summary bar (1) provides a visual summary of AI Evidence Analyzer results across all controls within the law or standard, while the AI Score column (2) shows the AI Evidence Analyzer result for each individual control.
This evaluation adjusts the control effectiveness score except for the not evaluated, such as controls that have no defined AM, controls with an answer of “No” or “N/A” when evidence is not required, and controls with no answer and also not in the state of Met by evidence or Partially Met.
You can also click on the result graphic to display more information.