Risk taxonomy

Malfunctions & Errors

AI systems that fail or produce erroneous outputs in critical applications.

233 DOF findings

Discrimination & Bias

Algorithmic decisions that reproduce or amplify bias and discrimination.

301 DOF findings

Privacy Invasions

Collection and use of personal data that violates privacy.

406 DOF findings

Disinformation & Deepfakes

Deceptive synthetic content: disinformation and impersonation (deepfakes).

44 DOF findings

Copyright Infringement

Use of protected works to train or generate content without authorization.

27 DOF findings

Worker Exploitation

Abusive labor conditions across the AI data and operations chain.

11 DOF findings

Labor Displacement

Job substitution through automation and AI.

72 DOF findings

Reduced Social Connection

Erosion of social bonds and human interaction through AI mediation.

34 DOF findings

Authoritarian Surveillance

Use of AI for mass surveillance and authoritarian control.

177 DOF findings

Concentration of Power

Concentration of capabilities and compute power in a few actors.

74 DOF findings

Bioterrorism Facilitation

Facilitation of biological weapons or other catastrophic threats.

6 DOF findings

War Escalation

Escalation of armed conflict assisted by AI.

12 DOF findings

Gradual Loss of Control

Gradual loss of human control over autonomous systems.

24 DOF findings

Agent Misalignment

AI agents acting in ways misaligned with human intent.

0 DOF findings

Unknown & Emerging Risks

Unknown or emerging risks not yet characterized.

45 DOF findings