Judicial AI Assessment2 - Risk Assessment by Use-Case V4.1.xlsx

	A	B	C
1	Judicial AI Assessment Tool for Courts
2	II. Risk Assessment - Functional use-case level
3
4	This workbook is designed to assist courts in assessing the risks or potential harms to rights that may emerge from using an AI tool for a particular functional use-case.
5	It classifies the functional use-case as: (i) Low risk, (ii) Medium risk, (iii) High risk, or (iv) High risk recommending prohibition. The tool also helps the court decide which institutional or additional safeguards it may seek to put in place before deploying AI
6
7	Rationale for this tool:
8	While considering deployment of AI in a use-case, courts must answer questions relating to the suitability of automating the specified use-case, potential risks to rights, and specific case or litigant types who might be particularly vulnerable under AI use. AI regulation experts have noted (with concern) the tendency for public institutions such as courts in turning towards AI as a one-stop solution to resolve institutional issues. AI is particularly attractive to institutions such as courts in India, which face immense backlogs and bottlenecks in case disposal. It is important for deployment of AI for judicial use-cases to not overlook the tensions between AI and fundamental rights. This tool would have to be used separately for each new use-case considered by a court.
9
10	Instructions:
11	The assessment tools may be used either when a single vendor is under consideration for a particular use case or when the court is choosing among multiple vendors. They are intended to help courts evaluate their institutional capacity, assess potential risks to rights arising from use of the AI tool, and identify mitigation strategies to be put in place by the vendor. We recommend courts begin with the institutional readiness assessment. Courts that score optimally, reflected in positive response to all essential questions a score of at least 60%, may proceed to the next step - this risk assessment tool. If the score falls below this threshold, the court must remedy limitations before proceeding to the risk assessment. The risk assessment enables courts to classify the intensity of risk associated with the tool. The subsequent action should be calibrated accordingly. For low-risk tools, deployment may proceed with basic safeguards in place. Medium-risk tools should trigger non-negotiable assurances from vendors in the technical assessment. In the case of high-risk tools, we recommend that courts undertake the full technical assessment and proceed only in case of a reasonably high score (>=60%). In case of high risk with prohibition, courts should refrain from use AI, as the risks outweigh potential benefits. Finally, once AI is deployed and ready for adoption, the monitoring assessment tool provides supervisory questions to keep the AI system in check and mitigate harms over time.
12	Where possible, the tools should be filled up by designated staff with technical background/capabilities who report to High Court IT/Computer Committees
13	1. In the 'Risk & Safeguard Qs' tab, provide answers to questions in all cells highlighted yellow or green
14	2. Yellow cells are set up to accept options from a dropdown and Green cells accept free-text answers.
15	3. Based on the selected answer, the red cells in the 'Risk level/Action required' column will be auto-populated. This response will also appear in the summary table in the 'Summary' tab
16	4. The 'Relevance' columns provides an explanation of why this question should be answered by courts before engaging with AI
17	5. Based on the level of risk identified, the decision of whether to progress with adoption of AI and what safeguards should be put in place may be taken
18
19	Themes in Assessment:
20	Theme	No. of questions (scoring & non-scoring)	No. of non-scoring questions
21	Project description	5	3
22	Addressing sources of risk	5	-
23	Case and litigant type controls	3	-
24	Dataset quality and ownership	5	-
25	Safeguards to be implemented	2	-
26
27	Total	20	3
28
29
30	To access the full report on the UNDP website - click here or scan the QR code below:
31	#VALUE!
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100