This Week in Email - April 15, 2026

The CNIL just published final rules requiring separate consent for email tracking pixels — the most consequential regulatory move for open-rate tracking since Apple MPP. The FBI's annual cybercrime report crossed $20.9 billion for the first time, with BEC still the second-costliest category. Apple launched its business email platform today. Cisco Talos found attackers weaponizing GitHub and Jira notification emails to bypass every gateway on the market. And Microsoft's authentication enforcement deadline is two weeks out.

Heavy week. Let's get into it.

Category: Regulatory

The CNIL — France's data protection authority — today published its final recommendation on tracking pixels in email, and it is a significant shift. Tracking pixels are now classified as equivalent to cookies under French law, meaning a single "I agree to receive marketing emails" checkbox no longer covers open-rate tracking. Organizations must obtain two independent consents: one for the marketing email itself, and a separate, specific consent for the tracking pixel.

There's a technically demanding wrinkle: when recipients withdraw consent, the change must take effect at the pixel server level immediately — including for pixels in emails already sitting in their inbox. That means real-time consent verification at pixel load time. Transactional emails and truly anonymous aggregate tracking (identical pixel, no per-recipient identifiers) are exempt. Existing senders have a 3-month transition window, roughly until July 14.

This is arguably the most important regulatory development for email open-rate tracking since Apple's Mail Privacy Protection in 2021. Between MPP, Gmail AI auto-opens, and now CNIL's double consent — open rate as a metric is getting squeezed from every direction. If your measurement strategy still leans heavily on opens, the time to shift toward clicks, conversions, and replies was yesterday.

Sources: CNIL Official | ChapsVision Analysis | Didomi Analysis | Badsender

Category: Security

The FBI's 2025 Internet Crime Report hit landmark numbers: over 1 million complaints for the first time, $20.88 billion in losses (up 26% from 2024). Business Email Compromise remained the second-costliest category at $3.05 billion across nearly 25,000 complaints. For the first time, the report breaks out AI-enabled fraud as a distinct category — 22,000+ complaints with $893 million in losses. BEC schemes with a confirmed AI component accounted for over $30 million alone.

Email remains the primary initial access vector across nearly every crime category the FBI tracks. The AI-enabled fraud number is almost certainly an undercount — detection attribution is still immature — but the fact that the FBI now categorizes it separately tells you where the threat landscape is heading.

Sources: IC3 Report (PDF) | The Register | SecureWorld

Category: Platforms

Apple Business went live this week in 200+ countries. The free platform includes business email, calendar, and directory services on custom domains for up to 500 users. It replaces Apple Business Essentials, Apple Business Manager, and Apple Business Connect — all discontinued. This is Apple's first direct play in business email hosting, putting them in competition with Google Workspace and Microsoft 365.

The catch: the email/calendar/directory features require iOS 26, iPadOS 26, or macOS 26. That's a meaningful ecosystem lock-in. If your organization runs mixed-OS or has devices that can't jump to the latest version, this isn't an option. For small Apple-native shops, though, a free business email suite with tight device integration is a compelling proposition. Worth watching how this affects the SMB email hosting landscape.

Sources: Apple Newsroom | TechSpot | PPC Land

Category: Security

Cisco Talos published research showing attackers embedding phishing links inside SaaS notification emails from GitHub and Jira. The emails pass SPF, DKIM, and DMARC because the platforms themselves send them. On GitHub, attackers drop malicious links into issue and PR descriptions that flow into notification emails. On Jira, they abuse the "Invite Customers" feature to inject phishing into Atlassian's own email templates. On peak days, roughly 2.89% of all GitHub emails were abuse-related.

This is a hard problem because there's nothing wrong with the emails at the authentication level. The sender infrastructure is legitimate. The message templates are legitimate. The content is the only signal — and it's user-generated content flowing through trusted pipes. Gateway-level defenses that rely on sender reputation and authentication status are essentially blind to this vector.

Sources: Cisco Talos | Help Net Security | GBHackers

Category: Deliverability

Sinch Mailgun's 2026 Email Impact Report, based on 400 billion+ emails and 1,200+ senders across 5 countries, puts a hard number on the inbox placement gap: nearly 18% of all emails fail to reach the inbox. And only 46% of organizations can actually measure their promotional email ROI — though among those who can, 60% report $10+ return per dollar spent.

The disconnect is striking: 78% say email is critical to their business, but fewer than half can prove it. The report also flags that AI adoption in email is still mostly limited to content generation — the teams applying AI to deliverability optimization, segmentation, and send-time decisions are seeing meaningfully better results, but they're in the minority.

Sources: Mailgun Report | MarTech | Telecom Reseller

Two weeks until the April 30 deadline. After that date, high-volume senders (5,000+ emails/day) to Outlook, Hotmail, and Microsoft 365 without SPF/DKIM/DMARC alignment will see emails junked or rejected. Compliance failures now include missing PTR records, lack of TLS, high complaint rates, and missing one-click unsubscribe. This completes the trifecta — Google (2024), Yahoo (2024), and now Microsoft all enforce strict authentication.

Sources: Microsoft Tech Community | Redsift Guide

Microsoft disclosed a campaign targeting Canadian workers via poisoned search results for "Office 365." Victims click top results that lead to adversary-in-the-middle credential harvesting pages. Once inside, the attackers change direct deposit details to redirect salary payments to accounts they control. Not industry-specific — pure geographic targeting. The payroll fraud vector is growing fast alongside generative AI that helps attackers identify payroll managers and craft impersonation messages.

Sources: Microsoft Security Blog | Help Net Security

IRONSCALES highlighted five attacks showing the shift from volume to precision: a zero-click PDF that auto-launched a credential harvest, a QR code phishing PDF with the target's email pre-encoded in base64, phishing kit assets hosted on a commercial phishing simulation vendor's S3 bucket, a Google Drive share notification used for law firm impersonation (passed SPF/DKIM/DMARC because Google sent it), and an invoice thread typosquat with a one-letter domain difference. Every attack was built for a specific person before it was sent.

Sources: IRONSCALES | Security Boulevard

Every email Postmark sends now runs through KumoMTA, the open-source Rust-based MTA. The key driver: PowerMTA's proprietary limitations and aging hardware made real-time traffic shaping difficult at scale. The performance numbers are notable — Gmail queue times dropped from ~2s to ~1.2s, Yahoo from ~4.6s to ~3.2s, Microsoft from ~4.8s to ~2.8s. Bounce rates held stable. Another major sender choosing open-source over proprietary is a meaningful signal for the commercial MTA market.

Sources: Postmark Blog | KumoMTA

Mailjet (Sinch) released an open-source MCP (Model Context Protocol) Server — a bridge that lets AI tools like Claude connect directly to Mailjet's API for contact management, campaigns, segmentation, and analytics. Built on Anthropic's open MCP standard. Read-only by default. This is one of the first ESPs to ship native MCP integration, and it signals a broader trend: ESPs building AI-native interfaces beyond "generate a subject line for me."

Sources: Mailjet Blog | GitHub

Folderly's analysis introduces "effective inbox placement" as a metric — and the numbers are alarming: up to 40% of emails that technically reach the Gmail inbox are being deprioritized by AI filtering. Not blocked. Not bounced. Just quietly buried. Gmail's Gemini-powered AI is evaluating content quality, structure, and value density as direct delivery signals. AI auto-opens for summarization are also inflating open rate metrics even further.

The recommendations: front-load key information in the first 100-200 characters, use clear structure, eliminate filler, deliver obvious value immediately. In other words, write emails worth reading. The AI is optimizing for the reader — and if your email doesn't hold up to that standard, it'll get deprioritized whether it "delivered" or not.

Sources: Folderly | Zepic