Forum Discussion

purplesnoble's avatar
purplesnoble
Copper Contributor
Oct 23, 2026

Outlook.com Spam Filtering Regression — Missing Rule Conditions Prevent Safe User‑Level Mitigation

I’m a program manager working with data engineers in a large corporation, and I’ve used Microsoft Exchange professionally for over 30 years. I’m posting because Outlook.com’s spam filtering pipeline appears to have regressed significantly, and the current rule system no longer provides the necessary conditions for users to mitigate the issue themselves.

 

Summary of the Issue

 

Over the past several months, Outlook.com has begun allowing large volumes of spam across multiple categories — explicit content, romance scams, foreign‑language marketing, calendar spam, health spam, lotto scams, and unsolicited newsletters. This is not typical spam leakage; it suggests a failure or degradation in one or more layers of the filtering pipeline (content classification, sender reputation scoring, behavioral heuristics, or transport‑level filtering).

 

Technical Problem

Outlook.com previously offered rule conditions that allowed users to safely filter spam based on message behavior and metadata. These included:

Message type (new vs. reply)

Does not include (critical for safe filtering)

Header includes

Unsubscribe detection

Cold sender / no prior conversation

These conditions are no longer available in the current Outlook.com rule interface.

The remaining conditions — primarily “includes” filters on subject, sender, or body — are insufficient for safe spam mitigation. Any attempt to delete spam using the current rule set risks deleting legitimate mail, because:

All legitimate transactional mail contains “http”

All receipts, confirmations, and alerts contain structured HTML

Many legitimate first‑contact messages contain links

There is no longer a way to exclude legitimate patterns (unsubscribe, structured headers, known senders, etc.)

 

Impact

Users cannot safely create rules that distinguish spam from legitimate mail. This leaves them with only two options:

Allow spam to reach the inbox

Move all messages containing “http” to Junk, which also captures legitimate mail

This is a regression in Outlook.com’s functionality and user safety. It also disproportionately affects users who rely on Outlook.com for critical communications.

 

Escalation

I have already escalated this issue directly to Microsoft leadership:

Satya Nadella (CEO)

Brad Smith (President)

Rajesh Jha (EVP, Experiences & Devices)

Steve Chew (Corporate VP, Outlook & Exchange)

Chris Capossela (Chief Customer Officer)

 

I’m posting here to ensure visibility within the Outlook engineering and product teams, and to request confirmation that:

 

This regression is known and being investigated

The missing rule conditions will be restored

Outlook.com’s spam filtering pipeline is being reviewed for failures in classification or reputation scoring

No part of this functionality will be moved behind a paid tier

 

Why This Matters

Outlook.com has historically been a reliable consumer‑grade and professional‑grade mail platform. Removing advanced rule conditions without providing alternative filtering controls leaves users unable to protect themselves from high‑volume spam. This is a user safety issue, not a feature request.

I’m asking the Outlook team to prioritize this regression and restore the rule capabilities that previously allowed safe and effective spam management.

 

Thank you for your attention.

2 Replies

  • Victor_Ivanidze's avatar
    Victor_Ivanidze
    Bronze Contributor

    The outlook.com is a free email service, correct? A bit strange to read your complaints then.

    • purplesnoble's avatar
      purplesnoble
      Copper Contributor

      My understanding was that this is a technical community. Have I mistaken it for a forum for critiquing the legitimacy of other users' questions?

      For context, I am a longstanding paying Microsoft 365 Personal subscriber. Outlook.com is where Microsoft directs me to manage certain settings.

      If you have a technical solution to the spam-filtering issue I've documented, I'd genuinely welcome your contribution.