1.2.0 - 2026-08-01

Added

• Opt-in on-device email learning for short local scam-pattern fingerprints
• Domain memory for custom From domains (not broad free webmail providers alone)
• In-message Not a scam control with a short undo window
• One-time inbox prompt and settings control for learning consent
• Post-update on-device notice when a new extension version and privacy changes ship
• Stronger inbox heuristics for suspicious webmail messages and links
• Stronger local detection for scrambled scam hosts and abuse-prone domain shapes
• Detection for fake support or report desktop-app download pressure
• Detection for fake Amazon, PayPal, and bank login pages on lookalike hosts
• Calm Possible scam highlights on Gmail, Outlook, Proton, Yahoo, and AOL webmail
• Hover reasons for webmail warnings instead of intrusive popup alerts

Changed

• Privacy policy revision 1.2.0 (effective August 1, 2026) documents webmail
  reading, email warnings, on-device learning, permissions, and retention
• Published privacy contact points to the Grandma Guard support form at
  https://nokaangel.dev/support?project=grandmaguard
• Public privacy page content aligned with the published policy

Fixed

• Replaced unsafe innerHTML assignments in detector.js and mail-guard.js
  with DOM createElement / textContent construction so Firefox AMO
  validation no longer flags those warnings

Security / privacy

• Still no telemetry, remote code, analytics, or off-device data collection

Current Issues
We're aware that sometimes the "Possible Scam" or "Likely Scam" highlighter will not show up properly, we are working on a fix for this.

Version 1.0.1 improves project transparency by providing the complete MIT-licensed source code and reproducible build instructions. Browser permissions, privacy practices, and protection behavior are unchanged. This update adds no telemetry, remote code, analytics, or data collection.