Back to the MIT repository
2. Privacy & Security3 - Other

Security

every piece of software, including learning systems, may be hacked by malicious users

Source: MIT AI Risk Repositorymit596

ENTITY

1 - Human

INTENT

1 - Intentional

TIMING

3 - Other

Risk ID

mit596

Domain lineage

2. Privacy & Security

186 mapped risks

2.2 > AI system security vulnerabilities and attacks

Mitigation strategy

1. Implement a rigorous AI supply chain security protocol, mandating the vetting and validation of all third-party datasets, models, and integrations to confirm security and integrity prior to deployment. 2. Apply layered security controls, including robust encryption for data both at rest and in transit, coupled with strict access controls following the principle of least privilege for sensitive data and model endpoints. 3. Establish a continuous adversarial testing regimen, such as ethical hacking, red teaming, and penetration testing, to proactively discover and mitigate AI-specific security vulnerabilities and attack surfaces.