Back to the MIT repository
7. AI System Safety, Failures, & Limitations2 - Post-deployment

Law abiding

We find literature that proposes [38] that early artificial intelligence should be built to be safe and lawabiding, and that later artificial intelligence (that which surpasses our own intelligence) must then respect the property and personal rights afforded to humans.

Source: MIT AI Risk Repositorymit113

ENTITY

2 - AI

INTENT

2 - Unintentional

TIMING

2 - Post-deployment

Risk ID

mit113

Domain lineage

7. AI System Safety, Failures, & Limitations

375 mapped risks

7.3 > Lack of capability or robustness

Mitigation strategy

1. Implement advanced AI alignment techniques, such as Constitutional AI or Cooperative Inverse Reinforcement Learning (CIRL), to formally encode and continuously maintain the system's objectives with human values and comprehensive legal/ethical constraints, specifically those related to personal and property rights, thereby proactively mitigating the risk of goal misalignment as capability increases. 2. Mandate the rigorous application of formal verification methods to core control and decision-making modules to mathematically prove that the system's design and execution rigorously satisfy the specified law-abiding and rights-respecting requirements, directly addressing the underlying "Lack of capability or robustness" sub-domain. 3. Establish a stringent governance framework, including continuous Human-in-the-Loop (HITL) oversight for high-impact decisions and transparent accountability protocols, ensuring the capacity for prompt human intervention, auditing, and redress of unintended consequences that may violate legal or personal rights.