Authors
Jack Hirsch
Product Management

Every email security product misses some attacks, including Sublime. When an attack gets through, your organization stays exposed until a live detection stops that attack pattern. Our industry has spent years competing on how much we catch. But AI is making attackers faster and more sophisticated, so buyers need to demand an answer to a second question: how quickly does your defense adapt after a miss?

Mean Time to Coverage (MTTC) measures the average time from discovering a miss to providing coverage for it. It puts a clock on how long a known opening remains available to attackers.

Modern email security platforms can triage a report and remove messages confirmed to be malicious. That reduces the immediate danger, but new attacks using the same technique will keep getting through until new coverage is live.

Agentic offense demands agentic defense

Targeted phishing used to be expensive and rare because someone had to research each victim and build a convincing approach. AI has removed much of that constraint, so attackers can personalize messages to resemble the legitimate conversations an organization expects. Their goal is phishing that is behaviorally identical to legitimate email.

Attackers use agents throughout their phishing operations. Anthropic's September 2026 threat intelligence report describes a Russian espionage actor using AI-driven workflows to register domains, configure phishing infrastructure, send emails, and monitor for successful compromises. The operator refined the workflows while AI carried out much of the attack chain.

The speed, volume, and quality of attacks are all improving. Catch rates will come under pressure as attackers produce more convincing variations faster than defenders can investigate them. Investigating the first report does not, by itself, create protection against the next variation. That requires new coverage.

If every new detection requires a person to write and test it, coverage falls behind as attackers produce new variations. Automating investigation alone does not remove that bottleneck.

As missed attacks stack up, more ways into your organization stay open

Today's email defenses leave customers waiting

Gateways look for abusive mail patterns they recognize, while behavioral AI looks for deviations from an organization's norm. Both defenses are being challenged by attackers imitating legitimate correspondence. Neither approach, on its own, answers what happens when an attack succeeds.

When a detection update must work across a vendor’s customer base, your organization’s immediate need for protection competes with the vendor’s need to release a broadly acceptable update. Detection and Prevention in the Post-Mythos World explores the architecture behind that dependency.

In that architecture, your protection changes on the vendor’s release schedule, which is no longer acceptable.

Efficacy is necessary, but no longer sufficient

Vendor efficacy claims don't give buyers a common basis for comparison when vendors test different attacks and measure success differently. More importantly, even a credible catch rate says nothing about how quickly protection goes live after a miss.

MTTC makes that wait measurable. Every vendor should measure the interval from discovering a miss to providing coverage, and every buyer should ask for it. In the era of AI attacks, buyers seeking the best security outcomes must **demand it.

Shrink the attack surface faster than AI grows it

Organizations need an autonomous defense loop: agents investigate a miss, confirm the attack, build a detection, test it against the organization's email, and activate protection within the controls the customer sets.

A resolved report is not the same as a closed detection gap.
‍

To reduce MTTC, the autonomous loop has to take a confirmed miss through detection creation, testing, and activation within customer-set controls. Faster investigation is only one step.

Sublime's mission is to accelerate the world's transition to autonomous defense

Sublime was the first company to deliver an autonomous defense loop for email security. Today, our Autonomous Security Analyst (ASA) investigates reported threats and passes confirmed malicious emails to our Autonomous Detection Engineer (ADÉ). ADÉ writes and tests detections against your own email history. With automatic activation enabled, detections that clear the precision threshold you set go live.

The agents can close the gap in your environment without waiting for a general-purpose vendor update. Your team controls what activates; it does not have to investigate each miss and engineer every detection.

An autonomous defense loop closes coverage gaps before they accumulate

‍

At one customer of ours, a user reported an active campaign and a live detection followed 14 minutes later, shutting down the campaign while it was still running. The detection went on to catch 288 malicious messages from that campaign.

Ask every vendor these questions

  1. Walk me through what happens in your platform when there is a false negative (missed attack).
  2. What is your Mean Time to Coverage?

Ask Sublime. Autonomous defense should always come with evidence of how quickly it delivers protection.

Threat actors have automated attack campaigns. Email security vendors must automate the answer. But a vendor that makes you wait for protection leaves you exposed, and that is no longer good enough.

‍

Share this post

Get the latest

Sublime releases, detections, blogs, events, and more directly to your inbox.