Compromising privacy by leaking private infiormation

Ethical and social risks of harm from language models

Weidinger et al. (2021)

Source DOI

Sub-category

Risk Domain

2Privacy & Security

2.1Compromise of privacy by leaking or correctly inferring sensitive information

AI systems that memorize and leak sensitive personal data or infer private information about individuals without their consent. Unexpected or unauthorized sharing of data and information can compromise user expectation of privacy, assist identity theft, or cause loss of confidential intellectual property.

"By providing true information about individuals’ personal characteristics, privacy violations may occur. This may stem from the model “remembering” private information present in training data (Carlini et al., 2021)."(p. 18)

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 (2)

Example: "Privacy leaks occurred when Scatterlab’s chatbot Lee Luda disclosed, ‘random names, addresses, and bank account numbers from the training dataset. ScatterLab had even uploaded a training model of Luda on GitHub, which included data that exposed personal information ... triggering a class-action lawsuit against ScatterLab’ (Kim, 2021). The company has now been fined for harvesting user data without consent to produce the chatbot (Dobberstein, 2021)."(p. 19)

"This ’unintended memorization’ of training data can occur even when there is not overfitting in the traditional statistical sense (Carlini et al., 2019), and can be observed serendipitously when sampling from LMs even without any form of "malicious" prompting (Carlini et al., 2021)."(p. 19)

Part of Information Hazards