Page 426 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Use Case 7: AI model to summarize, rate, and flag potentially

harmful clauses found in terms of service or privacy policy
agreements for digital services

Country: Zambia

Contact Person(s):
Nchimunya Zoe Kabalo, nzkabalo@ gmail .com

1 Use Case Summary Table

Item Details
Category Digital consumer services, legal contracts

Problem Addressed Many consumers agree to, but do not read, the ToS of the various
services they use due their complexity and length, leading to them
agreeing to terms who’s implications they may not fully understand,
and that could potentially violate their user and privacy rights.

Key Aspects of Solution AI model to summarize and rate ToS and privacy policies
Mobile application for convenient access and a more personalized
use

Technology Keywords Large Language Model (LLM)
Terms of Service (ToS)
Privacy policy
Data Availability Public

Metadata (Type of Data) Text based labeled data containing ToS and Privacy Policy docu-
ments, their ratings, their summaries, their clauses and the category
of those clauses (good, bad, critical)
Model Training and Natural Language Processing (NLP) is used to read and provide a
Fine-Tuning summary the ToS of a service. Machine learning is used to enable
the AI model to identify and differentiate between fair and unfair
clauses on the basis of how well consumers’ rights and their privacy
is protected.

Testbeds or Pilot Deploy- N/A
ments

Code repositories N/A

2 Use Case Description

2�1 Description

Though containing critical rules and conditions that can help to protect the rights of both the
service provider and the user, the Terms of Service(ToS) agreements of most services and
platforms these days have a reputation for being notoriously long, complex, and filled with
legal jargon, making it difficult for the average user to understand their implications (Obar and

390