BackStructured access refers to methods which limit users’ or deployers’ direct access to a model’s parameters by constraining access to a model through a centralized access point (e.g., an API) [88]. This access point can be monitored for usage, and access can be revoked to users or downstream deployers in cases of misuse [105]. Within this centralized access point, automated filtering-based monitoring can be done on both inputs and outputs to ensure the model’s intended use is preserved [36]. This filtering can sometimes be supplemented by human oversight, depending on desired robustness levels.

HomeMitigations V2BrowseStructured access refers to methods which limit users’ or deployers’ direct access to a model’s parameters by constraining access to a model through a centralized access point (e.g., an API) [88]. This access point can be monitored for usage, and access can be revoked to users or downstream deployers in cases of misuse [105]. Within this centralized access point, automated filtering-based monitoring can be done on both inputs and outputs to ensure the model’s intended use is preserved [36]. This filtering can sometimes be supplemented by human oversight, depending on desired robustness levels.

Structured access to a model

Impacts of AI

Structured access to a model

Impacts of AI

This page is still being polished. If you have thoughts, please share them via the feedback form.

Data on this page is preliminary and may change. Please do not share or cite these figures publicly.

Structured access refers to methods which limit users’ or deployers’ direct access to a model’s parameters by constraining access to a model through a centralized access point (e.g., an API) [88]. This access point can be monitored for usage, and access can be revoked to users or downstream deployers in cases of misuse [105]. Within this centralized access point, automated filtering-based monitoring can be done on both inputs and outputs to ensure the model’s intended use is preserved [36]. This filtering can sometimes be supplemented by human oversight, depending on desired robustness levels.

Gipiskis (2024)|LLM classified

Mitigation Taxonomy

1AI System

1.2Non-Model

1.2.2Runtime Environment

Containment, isolation, and control mechanisms for system execution.

Also in Non-Model

1.2.1 Guardrails & Filtering1.2.3 Monitoring & Detection1.2.4 Security Infrastructure1.2.5 Provenance & Watermarking

Definitionp. 44

AI systems can be developed and tested within a sandbox, (a secure and isolated environment used for separating running programs), such that outside access to information within the sandbox is restricted. Within this environment, resources such as storage and memory space, and network access, would be disallowed or heavily restricted [15]. With sandboxing, dangerous or harmful outputs generated during testing will be contained.

LLM Classification Details

Reasoning

Centralized API access point with automated input/output filtering monitors and controls model usage to prevent misuse.

Code: 1.2.1Version: v0.6Classified: Feb 6, 2026

Part of

Cybersecurity

Other mitigations from Gipiskis (2024) (112)

Model development

2.4 Engineering & Development

Lifecycle:Build and Use ModelActor:DeveloperAIRM:Unable to classify

Model development > Data-related

1.1 Model

Lifecycle:Collect and Process DataActor:Unable to classifyAIRM:Unable to classify

Model evaluations

2.2.2 Testing & Evaluation

Lifecycle:Verify and ValidateActor:DeveloperAIRM:Measure

Model evaluations > General evaluations

2.2.2 Testing & Evaluation

Lifecycle:Verify and ValidateActor:DeveloperAIRM:Measure

Model evaluations > Benchmarking

3.2.1 Benchmarks & Evaluation

Lifecycle:Other (outside lifecycle)Actor:DeveloperAIRM:Measure

Model evaluations > Red teaming

2.2.2 Testing & Evaluation

Lifecycle:Verify and ValidateActor:DeveloperAIRM:Measure

View all 112 mitigations from this source →

Source Document

Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems

Gipiškis, Rokas; San Joaquin, Ayrton; Chin, Ze Shen; Regenfuß, Adrian; Gil, Ariel; Holtman, Koen (2024)

Organizations and governments that develop, deploy, use, and govern AI must coordinate on effective risk mitigation. However, the landscape of AI risk mitigation frameworks is fragmented, uses inconsistent terminology, and has gaps in coverage. This paper introduces a preliminary AI Risk Mitigation Taxonomy to organize AI risk mitigations and provide a common frame of reference. The Taxonomy was developed through a rapid evidence scan of 13 AI risk mitigation frameworks published between 2023-2025, which were extracted into a living database of 831 distinct AI risk mitigations.

View source DOI: 10.48550/arXiv.2410.23472

Classification

AI Lifecycle Stage

Deploy

Releasing the AI system into a production environment

Verify and ValidateOperate and Monitor

Responsible Actor

Developer

Entity that creates, trains, or modifies the AI system

DeployerUser

NIST AI RMF Function

Manage

Prioritising, responding to, and mitigating AI risks

Measure

Risk Domains

Primary

4 Malicious Actors & Misuse

Other

7.2 AI possessing dangerous capabilities 2.2 AI system security vulnerabilities and attacks

Structured access refers to methods which limit users’ or deployers’ direct access to a model’s parameters by constraining access to a model through a centralized access point (e.g., an API) [88]. This access point can be monitored for usage, and access can be revoked to users or downstream deployers in cases of misuse [105]. Within this centralized access point, automated filtering-based monitoring can be done on both inputs and outputs to ensure the model’s intended use is preserved [36]. This filtering can sometimes be supplemented by human oversight, depending on desired robustness levels.

Gipiskis (2024)|LLM classified

Mitigation Taxonomy

1AI System

1.2Non-Model

1.2.2Runtime Environment

Containment, isolation, and control mechanisms for system execution.

Also in Non-Model

1.2.1 Guardrails & Filtering1.2.3 Monitoring & Detection1.2.4 Security Infrastructure1.2.5 Provenance & Watermarking

Definitionp. 44

LLM Classification Details

Reasoning

Centralized API access point with automated input/output filtering monitors and controls model usage to prevent misuse.

Code: 1.2.1Version: v0.6Classified: Feb 6, 2026

Source Document

Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems

Gipiškis, Rokas; San Joaquin, Ayrton; Chin, Ze Shen; Regenfuß, Adrian; Gil, Ariel; Holtman, Koen (2024)