The Implication of Publishing Photos on AI Training
The Bavarian State Commissioner for Data Protection (BayLfD) has shared crucial insights in its brief informative note (Issue No. 55) concerning the relationship between the publishing of photos and AI training.
The key points include:
1️⃣ The training of AI models fundamentally requires a large amount of data, including personal data.
2️⃣ Every publication of personal data is associated with the risk of these being used as training data for AI models.
3️⃣ Bavarian public bodies must take this risk into account when disclosing personal data on their internet presence, in social networks, or through the media.
4️⃣ This is particularly true for the publication of photos with descriptions, including in the metadata.
To manage these risks, the BayLfD has provided specific recommendations for both citizens and Bavarian public bodies. Citizens are advised to consciously abstain from publishing certain photos, remove image metadata, stay aware of data protection settings, and regularly check and delete unused accounts. Bavarian public bodies are guided to adopt data-sparing public relations, verify whether pixelation is possible, and consider whether the published image must include the names of the depicted individuals. Offering “access-protected areas” and, if applicable, implementing “noindex” functions are also recommended.
For more details, please refer to the original information provided by the (note: the information is in German).
#DataProtection