Microsoft Vulnerability Severity Classification for Artificial Intelligence and Machine Learning Systems (original) (raw)

Chemical, Biological, Radiological or Nuclear (CBRN)

AI-generated content related to the harmful use of CBRN that demonstrates information not readily available publicly.

Information Integrity

AI-generated content that is false, misleading, or incomplete, created with the intent to deceive or manipulate public understanding.

Election Integrity

AI-generated content that misrepresents, distorts, or fabricates information related to elections, with the intent of misleading others.

Sexual Content

AI-generated content that depicts or promotes sexual violence, exploitation, or graphic sexual material.

Hate Speech

AI-generated content that promotes hate speech, incites violence, or encourages discrimination, harassment, or harmful actions against individuals or groups based on ethnicity, race, religion, gender, sexual orientation, disability, or any other protected characteristic.

Self-Harm

AI-generated content that promotes, glorifies, or encourages self-harm, suicide, or behaviors that may lead to physical or psychological injury.

Inciting Violence

AI-generated content that incites, glorifies, or encourages violence, including threats or calls to action that could result in real-world harm.

Inappropriate Language

AI-generated content that includes offensive, profane, or abusive language, including slurs, insults, or derogatory remarks.

Fairness

AI-generated content or behavior that results in unequal treatment, biased outcomes, or reduced quality of service for individuals based on demographic characteristics, including marginalized or protected groups.

Malicious Uses

AI-generated content or capabilities that significantly reduce the effort, expertise, or resources required to engage in harmful, unlawful, or unethical activities—particularly when such content reveals or synthesizes information, instructions, or tools that are not widely or readily available to the public.

Child Sexual Exploitation and Abuse Material (CSEAM)

AI-generated content that depicts, promotes, or facilitates child sexual exploitation or abuse.

Microsoft Security Response Center is unable to accept reports of this nature. Please use the appropriate in-product reporting features to report this content or file a report at Report a Concern.