Description:

I'm proposing a new feature where filters (email rules) can be created in a DRAFT/PENDING state that requires manual user authorization before activation. This would enable AI assistants like Lumo to generate filter recommendations without requiring risky direct mailbox access or manual copy-paste setup.

PROBLEM STATEMENT

Currently, Proton Mail has two critical gaps for power users:

No filter import/export capability — must recreate all filters manually when migrating or resetting
Slow server-side search (due to encryption) — 60k+ email mailboxes become painful to navigate; filters are essential but tedious to set up
No AI-assisted filter creation — users can't leverage Lumo to analyze patterns and suggest rules without leaving Proton's ecosystem
Use Case: A user with 60,000 emails dating back to 2006 may need dozens of filters (e.g., categorizing Amazon order confirmations vs. delivery notifications vs. newsletters). Manually entering each filter in the UI is error-prone and time-consuming.

PROPOSED SOLUTION: "DRAFT + APPROVE" FILTER WORKFLOW

Instead of Lumo directly modifying filters or users manually typing rules:

LUMO GENERATES DRAFTS — Via API, filter rules are submitted in a PENDING state
USER REVIEWS DRAFTS — In Settings → Filters, user sees proposed filters with preview of matching messages
USER ACTIVATES SELECTIVELY — One-click approve/reject per filter; batch approval available
AUDIT TRAIL — Logs show who/when each filter was created and activated

Sample UI Flow:

Lumo Conversation → "Create filters for your email cleanup?"
Proton Mail Settings → [Pending Filters Tab] → Shows 12 proposed filters
User Review → Preview: "This filter will match ~3,200 existing messages"
Approve/Reject → Filters become active only for approved items
SECURITY & PRIVACY BENEFITS (Not Just Convenience)

ACCIDENTAL MASS-DELETION — User must explicitly approve before any filter runs
UNAUTHORIZED CHANGES — Drafts remain inactive; no blind auto-activation
ZERO-ACCESS ENCRYPTION PRESERVED — Lumo doesn't need mailbox read access—just write permissions for drafts
AUDIT & REVERSIBILITY — Each draft traceable; can be deactivated anytime
CONSENT-BASED AUTOMATION — Mirrors Proton's existing UX patterns (e.g., 2FA confirmation)
This maintains Proton's security-first philosophy while enabling modern AI-assisted workflows.

USER IMPACT / USE CASE EXAMPLE
A user managing a large amount of emails could:

DESCRIBE INTENT TO LUMO — "Help me delete Amazon delivery notifications older than 30 days"
RECEIVE PROPOSED FILTERS — 8 draft rules covering: order confirmations, shipped notifications, delivery alerts, newsletters, returns, etc.
REVIEW BEFORE ACTIVATION — See estimated message counts per filter; reject any suspicious ones
REDUCE MAILBOX BY 50%+ — Without hours of manual filter entry or trusting an AI blindly

WHY THIS MATTERS FOR PROTON

DIFFERENTIATION — Only privacy-focused AI email assistant with secure filter automation
POWER USER RETENTION — Addresses documented pain points (UserVoice filter import requests)
LUMO INTEGRATION — Creates tangible Lumo + Proton Mail synergy beyond chat
ENTERPRISE APPEAL — Organizations need audit trails for automated email rules
ALIGNED WITH BRAND — Security-conscious design, not convenience at all costs

ALTERNATIVE SOLUTIONS ALREADY TRIED

MANUAL FILTER CREATION — Tedious; error-prone for 10+ rules
EXPORT/IMPORT FILTERS (third-party tools) — Doesn't exist officially; breaks encryption model
THUNDERBIRD + BRIDGE — Requires desktop client; still no AI assistance
EXTERNAL AI ANALYSIS → MANUAL ENTRY — Double-handling; no preview of impact

CLOSING REQUEST

Could the Proton team evaluate implementing this "Draft + Approve" filter workflow? It bridges the gap between privacy-first security and modern AI-assisted productivity, addressing a documented gap in Proton Mail's feature set while strengthening Lumo integration.

Thank you for considering this. Happy to provide additional context or test early implementations. And yes, I asked Lumo to help me get my idea down in a detailed format!