What kinds of data and filesets can TeraDact desensitize for AI training?

TeraDact can prepare training data across all the formats your models may need:

• Structured datasets: Anonymize or tokenize fields in CSV, JSON, SQL exports, and database tables — preserving schema, data types, and statistical distributions while removing identifying values
• Documents and text corpora: Strip PII from Word documents, PDFs, emails, and plain text files used to train NLP, summarization, classification, or retrieval models
• Images: Remove faces, names visible in images, and sensitive visual content from image datasets for computer vision model training
• Audio transcripts and recordings: Remove speaker identification and spoken PII from audio used to train speech recognition, sentiment analysis, or voice models
• Video datasets: Desensitize video content used for object detection, activity recognition, or multimodal AI models