Page 69 - AI Standards for Global Impact: From Governance to Action

AI Standards for Global Impact: From Governance to Action

disinformation as part of orchestrated campaigns intended to spread untruths. The paper
provides definitions for misinformation, disinformation and malinformation as follows:

• Misinformation refers to false information but is not created or shared with the intention
of causing harm. 7 Part 2: Thematic AI
• Disinformation is false content intentionally created and disseminated to mislead, harm,
or manipulate.
• Malinformation is factual information used out of context with the intent to cause harm.
For example, publishing private data with malicious intent (e.g. revenge porn or non-
consensual intimate imagery) or altering contextual metadata to mislead.
Types of misinformation, disinformation, and malinformation vary considerably. The table below
from the paper provides some examples.

Table 1: Types of misinformation, disinformation, and malinformation

8
Fabricated content Usually, 100 % false and designed to deceive and do harm. Distinguishing between
the real and fabricated content is extremely difficult. Exposure to sophisticated
deepfakes used to promote fabricated content can deeply impact trust in the
messages citizens receive.
Manipulated content Genuine information or imagery that has been distorted. These types of content
often manipulate genuine content by doctoring an image, or use sensational head-
lines or click bait.

Imposter content Impersonation of genuine sources, very often using the branding of an established
agency or a reputable news agency. This form of disinformation takes advantage
of the trust people have in a specific organization, a brand or even in a person.
Adversaries will use phishing and smishing messages using a well-known brand
in an attempt to create an impression that the recipient(s) are receiving legitimate
content.

Misleading content Misleading information is created by reframing stories in headlines. This typically
uses fragments of quotes to support a wider point, often citing statistics in a way that
aligns with a position. Alternatively, it can be the deliberate decision not to cover
something because it undermines an argument. When making a point, everyone
is prone to drawing out content that supports their overall argument.
False context Factually accurate content combined with false contextual information, such as the
headline of an article failing to reflect the content. Basically, the genuine content
has been reframed. False context images are a low-tech but still a powerful form
of misinformation and disinformation.

Satire and parody Humorous but false stores passed off as true; there is no intention to harm, but
readers may be fooled. What was once treated as a form of art, is now vigorously
used to intentionally spread rumours and conspiracies. It is difficult to police as
the perpetrators argue they are merely doing something that shouldn’t be treated
seriously or literally. The danger of this type of misinformation and disinformation
is in the method and speed with which it gets re-shared. In doing so it is often
reshaped or reframed and a wider audience loses the connection with the original
messenger, failing to understand it as satire.

Source: Policy Paper

7 See https:// webarchive .unesco .org/ web/ 20230926213448/ https:// en .unesco .org/ fightfakenews, or non-
consensual
8 This type uses false content such as the example of a deepfake audio clip of London mayor Sadiq Khan
that was widely circulated on social media in November 2023. The actors used a simulation of the mayor’s
voice allegedly calling for pro-Palestinian marches to take precedence over Remembrance weekend
commemorations on the same day.

57