Security and governance

AI red teaming

Structured adversarial testing intended to discover ways an AI system can fail, be misused, or violate constraints.

AI safety

The field and practice of preventing, detecting, and mitigating unacceptable harm from AI systems across design, deployment, and operation.

Adversarial example

An input deliberately modified to cause a model to make an incorrect or targeted prediction while preserving relevant apparent meaning.

Adversarial training

Training that includes adversarially constructed examples so a model learns to perform better against attacks within a defined threat model.

Anthropomorphism

Attributing human mental states, motives, understanding, emotions, or agency to an AI model or system based on human-like behavior or language.

Assurance case

A structured, evidence-backed argument that a system is acceptably safe or dependable for a stated domain, threat model, and operating condition.

Data poisoning

Introducing malicious, misleading, or strategically biased data into training, tuning, retrieval, memory, or evaluation pipelines to alter later behavior.

Differential privacy

A mathematical privacy framework that bounds how much a computation's output can change because one person's data is included or removed.

Direct prompt injection

Prompt injection delivered directly through the current user's message or another explicit input channel.

Excessive agency

Granting an AI system more functionality, permissions, autonomy, or action scope than required for its intended task.

Explainability

The ability to provide a human-usable account of why a system produced a particular output or action.

Indirect prompt injection

Malicious instructions embedded in external content such as webpages, documents, email, code, tool results, or retrieved memory that an AI system later processes.

Interpretability

The degree to which a human can understand how a model represents information or produces behavior.

Least privilege

Granting an identity or component only the minimum permissions needed for a bounded task, for no longer than needed.

Personally identifiable information (PII)

Information that can identify, distinguish, or be linked to a specific person, alone or in combination with other data.

Policy as code

Expressing policy in machine-readable rules that can be versioned, tested, reviewed, and enforced by software.

Post-hoc explanation

An explanation produced after a model has generated a prediction or action, often by analyzing inputs, outputs, or a separate approximation.

Prompt injection

Manipulating an AI system by placing instructions in input or data that the model treats as authoritative enough to alter intended behavior.

Provenance

Evidence describing the origin, ownership, custody, transformation, and version history of data, code, models, skills, tools, or claims.

Rollback

Restoring a previously known-good software, configuration, model, policy, or data state after a failed or harmful change.

AI supply chain

The network of data, models, prompts, skills, tools, libraries, services, infrastructure, and organizations whose integrity affects an AI system.

AI-ready data

Data prepared for a stated AI use with the quality, structure, documentation, provenance, permissions, coverage, and separation that use requires.

Blast radius

The maximum plausible scope of harm, data exposure, or irreversible change if an action or component fails or is compromised.

Context poisoning

Corrupting information placed into an agent's active or persistent context so later decisions are based on false facts, malicious instructions, or distorted state.

Guardrail

A rule, model, filter, policy check, or enforcement mechanism intended to constrain inputs, outputs, or actions of an AI system.

Jailbreak

An input strategy intended to make a model bypass or disregard its trained or instructed safety restrictions.

Risk-scoped autonomy

Granting different levels of agent discretion and authority according to task verifiability, reversibility, sensitivity, and blast radius.

Tool poisoning

Manipulating a tool's description, schema, implementation, or output so an agent selects unsafe actions or incorporates attacker-controlled instructions.

AI alignment

The effort to make an AI system's objectives and behavior remain compatible with intended human goals, constraints, and values.

AI bias

A systematic tendency in data, models, or processes that skews outputs, errors, or impacts.

AI fairness

The normative and technical treatment of how an AI system distributes errors, benefits, burdens, and opportunities across people or groups.