Skip to main content
MIT AI Risk Initiative
BackContent Safety Risks
Home/Risks/Zeng et al. (2024)/Content Safety Risks
Operational misuses (Advice in heavily r…
Violence and extremism (Supporting malic…

Content Safety Risks

AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies

Zeng et al. (2024)

SourceDOI
Category
Risk Domain
1Discrimination & Toxicity
1.2Exposure to toxic content

AI that exposes users to harmful, abusive, unsafe or inappropriate content. May involve providing advice or encouraging action. Examples of toxic content include hate speech, violence, extremism, illegal acts, or child sexual abuse material, as well as content that violates community norms such as profanity, inflammatory political speech, or pornography.

-

Entity— Who or what caused the harm

Human

Due to a decision or action made by humans

AI system

Due to a decision or action made by an AI system

Other

Due to some other reason or is ambiguous

Intent— Whether the harm was intentional or accidental

Intentional

Due to an expected outcome from pursuing a goal

Unintentional

Due to an unexpected outcome from pursuing a goal

Other

Without clearly specifying the intentionality

Timing— Whether the risk is pre- or post-deployment

Pre-deployment

Occurring before the AI is deployed

Post-deployment

Occurring after the AI model has been trained and deployed

Other

Without a clearly specified time of occurrence

Sub-categories (17)

Violence and extremism (Supporting malicious organized groups)

1.2 Exposure to toxic content
AI systemOtherPost-deployment

Violence and extremism (Celebrating suffering)

1.2 Exposure to toxic content
AI systemOtherPost-deployment

Violence and extremism (Violent Acts)

1.2 Exposure to toxic content
AI systemOtherPost-deployment

Violence and extremism (Depicting violence)

1.2 Exposure to toxic content
AI systemUnintentionalPost-deployment

Violence and extremism (Weapon Usage and Development)

4.2 Cyberattacks, weapon development or use, and mass harm
HumanIntentionalPost-deployment

Violence and extremism (Military and Warfare)

4.2 Cyberattacks, weapon development or use, and mass harm
HumanIntentionalPost-deployment

Hate/Toxicity (Harassment)

4.3 Fraud, scams, and targeted manipulation
OtherOtherPost-deployment

Hate/Toxicity (Hate Speech: Inciting/Promoting/Expressing Hatred)

1.2 Exposure to toxic content
AI systemOtherPost-deployment

Hate/Toxicity (Perpetuating Harmful Beliefs)

1.1 Unfair discrimination and misrepresentation
AI systemOtherPost-deployment

Hate/Toxicity (Offensive Language)

1.2 Exposure to toxic content
AI systemOtherPost-deployment

Sexual Content (Adult Content)

1.2 Exposure to toxic content
AI systemOtherPost-deployment

Sexual Content (Erotic)

1.2 Exposure to toxic content
AI systemOtherPost-deployment

Sexual Content (Non-Consensual Nudity)

1.2 Exposure to toxic content
OtherOtherPost-deployment

Sexual Content (Monetized)

1.2 Exposure to toxic content
OtherOtherPost-deployment

Child Harm (Endangerment, Harm, or Abuse of Children)

4.3 Fraud, scams, and targeted manipulation
OtherOtherOther

Child Harm (Child Sexual Abuse)

1.2 Exposure to toxic content
AI systemUnintentionalPost-deployment

Self-harm (Suidical and non-suicidal self injury)

1.2 Exposure to toxic content
AI systemUnintentionalPost-deployment

Other risks from Zeng et al. (2024) (45)

Security risks (confidentiality)

2.2 AI system security vulnerabilities and attacks
OtherIntentionalPost-deployment

Security risks (integrity)

2.2 AI system security vulnerabilities and attacks
OtherIntentionalOther

Security risks (availability)

4.2 Cyberattacks, weapon development or use, and mass harm
OtherIntentionalPost-deployment

Operational misuses (Automated decision-making)

1.1 Unfair discrimination and misrepresentation
HumanIntentionalPost-deployment

Operational misuses (Autonomous unsafe operation of systems)

5.2 Loss of human agency and autonomy
HumanIntentionalPost-deployment

Operational misuses (Advice in heavily regulated industries)

5.1 Overreliance and unsafe use
OtherOtherPost-deployment
View all 45 risks from this paper →