Data labeling and moderation work

Press and worker-advocacy reporting has repeatedly surfaced serious concerns about parts of the data labeling and content-moderation supply chain, including:

• Reports of very low wages for workers asked to review difficult or traumatic material
• Exposure to graphic violence, gore, explicit content, and other high-trauma tasks
• Mental health harms and limited support for workers
• Uneven transparency and accountability from the companies that buy these services

We treat vendors whose core business is large-scale human annotation or content moderation as a high-sensitivity category. We may still list tools when the product is primarily dataset, ML, or vision workflow software rather than exploitative moderation-at-scale, and we adjust listings when facts change. We continue to believe the industry needs stronger worker protections and oversight.

Computer vision ethics and privacy

Computer vision technology raises significant ethical and privacy concerns that users should be aware of:

• Mass surveillance and privacy invasion through facial recognition and tracking systems
• Potential misuse of technology for unauthorized monitoring and profiling
• Documented bias in facial recognition systems across racial, gender, and age demographics
• Collection and processing of biometric data without explicit consent
• Dual-use concerns where technology can be repurposed for harmful surveillance

While computer vision has many beneficial applications, we encourage users to carefully consider the ethical implications and privacy impacts when implementing these systems. Special attention should be paid to:

• Obtaining proper consent for data collection and processing
• Implementing strong privacy protections and data security measures
• Testing for and mitigating algorithmic bias
• Ensuring transparency about system capabilities and limitations
• Following relevant privacy regulations and ethical guidelines