Page 72 - AI Standards for Global Impact: From Governance to Action

AI Standards for Global Impact: From Governance to Action

Table 2: Applying PDR framework to build trust in multimedia authenticity

Policy
Approach Method Benefit and/or outcome
Requirements
Transparency Labelling Informs users about various aspects
of the content. Clearly identifying if
the content was AI generated.

Watermarking Non-human perceptible markings
applied to content that provide infor-
mation about it.
Traceability Content Enables providing information about
provenance tools the content’s origin and changes to
establish accountability and attribu-
tion.

Prevention Accountability Conduct risk Enforcement can be made more
assessment efficient when areas are identified
as high risk. Prevalent abuse or
patterns of behaviour are identified
and treated as priorities. This proac-
tive approach helps mitigate the
risks associated with manipulated
content, ensuring that users are
protected from misinformation and
fraudulent activities.

User education Public awareness Reduces accidental misuse through
initiatives education about copyright laws and
the consequences of infringement.

Detection Detecting manip- Technological These solutions offer numerous
ulated content solutions benefits such as protecting intellec-
and deepfakes tual property, verifying image, audio,
text and video authenticity, and
aiding in online safety and security.
However, it creates a ‘back and forth
war’ with bad actors who attempt to
avoid these detectors.
For example: https:// arxiv .org/ abs/
2504 .2148
Data privacy Data handling and All data processed are subject to
adherence to data randomized manual review, ensuring
protection accuracy and compliance with data
legislations protection legislation.

60