Overview
- Chrome began rolling out the new site warnings for users of Safe Browsing’s Enhanced Protection on Tuesday, showing a darkened pop-up overlay for lower‑confidence 'suspicious' pages and keeping full‑page red interstitials for high‑risk 'dangerous' sites.
- Google moved Gemini‑powered scam detection from on‑device models to a more capable server‑side model for Enhanced Protection users, sending page text to the model only when sites show strong scam signals and combining that output with domain reputation to reach a verdict.
- When page content is uploaded to Google’s servers for analysis the company says it is processed ephemerally and is not logged, retained, or used to train models.
- Chrome now detects ClickFix social‑engineering tricks by scrubbing personally identifiable information from copied clipboard commands, sending suspicious payloads to Safe Browsing for analysis, and warning users before they paste or run dangerous commands.
- The rollout also turns on a default prompt that asks for approval before connecting to unencrypted HTTP sites, and Google says Enhanced Protection — introduced in 2020 and used by over 1 billion people by 2025 — will include a 'Mark as safe' option to reduce false positives and lessen disruption to legitimate sites and developer workflows.