Gmail and Outlook capture now picks up inline images from the email body, so emails that put their content inside images (newsletters, promotional emails) get summarized properly instead of looking empty. Tracking pixels and tiny icons are filtered out, and up to 10 of the largest images are sent along with the email text.

Phishing analysis now works on many more webmail providers, including Fastmail, iCloud Mail, Zoho Mail, GMX, mail.com, Tutanota, Yandex, Mail.ru, HEY, Posteo, Mailbox.org, Disroot, and dozens of regional providers (WP, Onet, Seznam, T-Online, La Poste, Orange, Libero, QQ, 163, Naver, Daum, and cPanel webmail on Gandi, Aruba, IONOS, Bluehost, Dreamhost, HostGator, Namecheap, and others). Added a generic email extractor that pulls subject, from, to, cc, date, body, links, and attachments on these hosts. Also improved primary image detection on pages that do not set og:image or twitter:image tags: now checks JSON-LD structured data and scans the main content area for the largest qualifying image, with filtering for avatars, icons, sprites, logos, and tracking pixels. No change to permissions or data collection.