1 | Scope of the baseline requirements | 0 |
3 | Terms and abbreviations | 0 |
4 | Stakeholder roles across the AI supply chain | 0 |
5.1.1-1 | AI security content in the cyber security training programme | 0 |
5.1.1-2 | Ongoing staff awareness of AI threats, with secure AI coding training for developers | 0 |
5.1.2-1 | Assess the need, the risks and the mitigations before building an AI system | 0 |
5.1.2-2 | Design for resilience against adversarial attack, unexpected input and failure | 0 |
5.1.2-3 | Create an audit trail for the AI system | 0 |
5.1.2-4 | AI-specific risk assessment and due diligence for external components | 0 |
5.1.2-5 | Match intended use to data sensitivity, and encourage staff to report AI security risks | 0 |
5.1.2-6 | Least-privilege, risk-assessed permissions for the AI system on other systems | 0 |
5.1.2-7 | Due diligence on external providers | 0 |
5.1.3-1 | Threat modelling and risk management for AI systems | 0 |
5.1.3-2 | Pass unresolved AI threats down the chain to operators and end-users | 0 |
5.1.3-3 | Assurance over external parties responsible for AI risks | 0 |
5.1.3-4 | Continuous monitoring and review of system infrastructure against risk appetite | 0 |
5.1.4-1 | Build in and maintain capabilities for human oversight | 0 |
5.1.4-2 | Make outputs easy for responsible humans to assess | 0 |
5.1.4-3 | Technical measures where human oversight is a risk control | 0 |
5.1.4-4 | Verify the data custodian's specified controls are implemented | 0 |
5.1.4-5 | Tell end-users which uses of the system are prohibited | 0 |
5.2.1-1 | Comprehensive asset inventory including interdependencies | 0 |
5.2.1-2 | Processes and tools to track, authenticate, version and secure AI assets | 0 |
5.2.1-3 | Disaster recovery plans tailored to attacks on AI systems | 0 |
5.2.1-4 | Protect sensitive data and sanitise inputs, with proportionate protection of weights | 0 |
5.2.2-1 | Evaluate access control frameworks for APIs, models, data and pipelines | 0 |
5.2.2-2 | Controls on externally offered APIs against attacks on the AI system | 0 |
5.2.2-3 | Dedicated, segregated environments for development and model tuning | 0 |
5.2.2-4 | Publish a clear vulnerability disclosure policy | 0 |
5.2.2-5 | AI system incident management plan and recovery plan | 0 |
5.2.2-6 | Cloud service contracts that support these requirements | 0 |
5.2.3-1 | Secure software supply chain processes for AI development | 0 |
5.2.3-2 | Justify, mitigate and disclose use of poorly documented or unsecured components | 0 |
5.2.3-3 | Re-run evaluations on released models before use | 0 |
5.2.3-4 | Tell end-users before models are updated | 0 |
5.2.4-1 | Document system design and maintenance plans, and publish component hashes | 0 |
5.2.4-2 | Record the provenance and use of public training data | 0 |
5.2.4-3 | Audit log of changes to system prompts and model configuration | 0 |
5.2.5-1 | Security assessment testing before release of models, applications and systems | 0 |
5.2.5-2 | Operator testing before deployment, using independent skilled testers | 0 |
5.2.5-3 | Share testing and evaluation findings with system operators | 0 |
5.2.5-4 | Evaluate outputs for leakage of model internals and unintended influence | 0 |
5.3.1-1 | Tell end-users where and how their data is used, accessed and stored | 0 |
5.3.1-2 | Accessible guidance for end-users, including limitations and security updates | 0 |
5.3.1-3 | Support end-users and affected entities through incidents | 0 |
5.4.1-1 | Provide security updates and patches, with contingency where updates are impossible | 0 |
5.4.1-2 | Treat major updates as a new version and support operators through model changes | 0 |
5.4.2-1 | Log system and user actions | 0 |
5.4.2-2 | Analyse logs for anomalies, breaches and unexpected behaviour | 0 |
5.4.2-3 | Monitor internal states of the AI system where useful for security | 0 |
5.4.2-4 | Monitor model and system performance over time for security-relevant change | 0 |
5.5.1-1 | Secure disposal when training data or a model is transferred | 0 |
5.5.1-2 | Secure deletion when a model or system is decommissioned | 0 |