Protegrity releases open-weights 278M PII detection model and benchmark

Protegrity released an open-weights 278M PII detection model, a RoBERTa fine-tune trained on 1.2B synthetic PII-bearing samples, plus the PII-SCOPE benchmark; model and paper are public.

• Axion 1.0 open agent foundation model released2026-11-04
• Palette fine-tunes 20B model to build decks in brand design language2026-11-03
• Transformers adds small model fixtures to catch regressions2026-11-03
• Hugging Face releases FineDNA dataset to teach LLMs DNA2026-11-01
• Security Affairs releases AI-cybersecurity newsletter round 62026-11-01