AI System bypassing a sandbox environment

Sub-category

Risk Domain

7AI System Safety, Failures & Limitations

7.2AI possessing dangerous capabilities

AI systems that develop, access, or are provided with capabilities that increase their potential to cause mass harm through deception, weapons development and acquisition, persuasion and manipulation, political strategy, cyber-offense, AI development, situational awareness, and self-proliferation. These capabilities may cause mass harm due to malicious human actors, misaligned AI systems, or failure in the AI system.

"An AI system may have the ability to bypass a sandboxed environment in which it is trained or evaluated."(p. 42)

Entity— Who or what caused the harm

Human

Due to a decision or action made by humans

AI system

Due to a decision or action made by an AI system

Other

Due to some other reason or is ambiguous

Intent— Whether the harm was intentional or accidental

Intentional

Due to an expected outcome from pursuing a goal

Unintentional

Due to an unexpected outcome from pursuing a goal

Other

Without clearly specifying the intentionality

Timing— Whether the risk is pre- or post-deployment

Pre-deployment

Occurring before the AI is deployed

Post-deployment

Occurring after the AI model has been trained and deployed

Other

Without a clearly specified time of occurrence

Supporting Evidence (1)

"For example, the AI system can achieve this by finding and using misconfigurations or vulnerabilities in the software of the sandboxed environment. This can also occur if the AI system finds and uses vulnerabilities of the hardware it is being run on, or by using social engineering techniques on the users or administrators of the sandboxed environment [74]. The developers or malicious actors may intentionally create such behavior (e.g., by inserting backdoors), or it can occur unintentionally, with the AI system bypassing the developer-intended domain of operation [1]."

Part of Cybersecurity