Photos Of Australian Kids Have Been Found In A Massive AI Training Data Set. What Can We Do?
A recent report reveals that AI models were trained on millions of images from the public internet, including photos of Australian children, without their consent. This discovery highlights a critical conflict between rapid technological advancement and established privacy protections. The case demonstrates that the availability of data online does not exempt developers from legal obligations, challenging the misconception that public accessibility equates to open permission for commercial use. This situation is directly relevant to open data because it underscores the urgent need to distinguish between open accessibility and open usability. While data may be publicly accessible, using it for AI training often involves scraping and processing personal information in ways that violate privacy laws, as seen in previous rulings against facial recognition firms. The inability of current AI models to "unlearn" specific data points means that privacy breaches are often permanent, emphasizing that open data principles must be balanced with strict ethical and legal guardrails to protect individual rights. The article argues for stronger enforcement of privacy laws and the implementation of specific codes for children’s data protection. It shifts the burden of responsibility away from parents, who face impossible odds in preventing their children’s images from being scraped, and places it squarely on tech companies. Ultimately, the piece calls for accountability, urging regulators to impose significant penalties on organizations that misuse personal data, ensuring that the drive for open data does not come at the cost of fundamental human privacy.
Source: menafn.comPublished on 2024-07-04
Related news
- Photos of Australian children found in AI training dataset, create deepfake risk | Biometric Update
- Regulaciones impulsan la implementación de modelos de gobernanza de datos | Diario Financiero
- Expertos en ciberseguridad desmontan el ‘pajaporte’: “es el sueño húmedo de un dictador”
- Even staunch fans are calling out Apple's less-than-transparent AI training data harvesting
- Bharat needs culturally aligned, multimodal AI models, not Western-centric ones: NVIDIA exe - OrissaPOST