Microsoft's July 2026 release notes for Security Exposure Management have added a long-anticipated feature: Codename MDASH, an agentic code scanner that uses a multi-model AI system to hunt for vulnerabilities. But the feature is labeled private preview, not the public or even expanded preview many organizations expected after Microsoft's June messaging. That gap puts security teams in a bind—they need to know whether they can get their hands on the scanner now, and if so, how to test it without stumbling into data-governance landmines.

What’s actually in the release—and what’s missing

On paper, the July 2026 update to Microsoft Security Exposure Management (MSEM) introduces MDASH as a working feature. According to Microsoft’s official documentation, security teams can launch scans from the Defender CLI or through a GitHub connector, then triage findings inside the Microsoft Defender portal. The scanner is designed to deliver “greater depth and accuracy than traditional static analysis” by orchestrating over 100 specialized AI agents that debate and validate potential flaws rather than just pattern-matching against a rule book.

That promise is real, and it’s ambitious. But the release notes stop short of answering the questions every CISO needs answered: Is my tenant eligible? How do I enroll? What will it cost? Where does my code get processed, and for how long? Microsoft hasn’t published separate pricing or a clear enrollment path. Even the preview label is contradictory. A May 12, 2026, announcement described MDASH as being tested with a “limited group of private-preview customers.” By June, the company was talking about an expanded preview for eligible organizations. Now the July notes flatly call it “private preview.” That whipsaw leaves IT leaders guessing whether they’re in or out—and whether the feature is actually production-ready enough to test at all.

What this means for you

If you’re a home user or small-business owner who doesn’t run a security operations center, this announcement won’t change your day-to-day. MDASH lives inside the enterprise-grade Microsoft Defender portal, and you’ll need a qualifying license—like Microsoft 365 E5, Defender for Endpoint, or Microsoft 365 Business Premium—to even see the option.

For security architects and admins, the impact is immediate: your evaluation timeline just got fuzzy. You likely carved out Q3 to trial MDASH alongside your existing static application security testing (SAST), dependency scanning, and secret-scanning tools. Now you must first figure out if the feature is actually visible in your tenant. Even if it is, a “private preview” label implies that Microsoft hasn’t finalized terms of use, data handling, or support. That’s a red flag for any production repository. You shouldn’t be pointing MDASH at your crown jewels until those gaps close.

Developers and DevOps teams need to hear a different message: a new scanner will eventually reach your CI/CD pipeline. Start mapping which repositories you’d be willing to test with, and make sure the owners of those services are ready to triage findings. The scanner promises to find flaws traditional tools miss, but if your dev backlog is already full, a flood of new alerts won’t improve security—it’ll just create noise.

Compliance officers and privacy leads face the tallest hurdle. MDASH ingests source code, which can contain proprietary algorithms, hard-coded infrastructure secrets, and internal service blueprints. Before any scan runs, you need documented answers about where code is uploaded, whether it’s used to train models, how long artifacts persist, and which geographies handle the data. Right now, Microsoft’s public material doesn’t answer these questions. Until you get those answers through contractual channels, treat MDASH as a high-risk data processor.

How we got here: a timeline of mixed signals

Microsoft’s path to exposing MDASH has been rapid—and confusing.

  • May 12, 2026: In a security blog post, Microsoft first unveiled an agentic vulnerability-discovery system. The company framed it as a private preview, with a select group testing the multi-model pipeline.
  • June 2026: At an Asia-focused event and in community channels, Microsoft discussed expanding the preview to more organizations. No firm date was set, but the language suggested imminent broader availability.
  • July 2026: The Security Exposure Management what’s-new page quietly listed MDASH as a “private preview” without fanfare. No separate blog post, no pricing page, no enrollment button—just a feature entry tucked into release notes.

This isn’t the first time Microsoft has swung between “private” and “public” for a high-profile security feature. The Defender for Cloud agentless code scanner, for instance, rolled out region by region over months. But MDASH’s appetite for raw source code makes a vague preview status riskier. Organizations that mistake the feature’s presence in the Defender portal for a green light could inadvertently violate data-residency rules or intellectual-property agreements.

What to do now: a practical playbook

If you plan to evaluate MDASH—now or once it opens wider—take these steps immediately. They’re designed to let you test safely without assuming anything Microsoft hasn’t confirmed.

1. Verify tenant availability

Log in to the Defender portal and navigate to Security Exposure Management. If you see MDASH options under scan configuration, you’re in. If not, don’t hammer support. The feature is still gated, and no amount of licensing qualifies you automatically. Ask your Microsoft account team for the current status and a timeline.

2. Pick a low-risk, well-understood service

Choose a non-critical codebase whose owners can review findings promptly. Avoid identity systems, signing infrastructure, payment modules, shared libraries, or any repos containing sensitive intellectual property. A small internal tool or a deprecated service branch is ideal.

3. Lock down access—tightly

Whichever path you use (GitHub connector or Defender CLI), create a dedicated identity with least privilege permissions. Grant read-only access only to the specific repository and branches needed for the trial. Do not allow organization-wide visibility. Document the identity and set an expiration date; revoke access when the test ends.

4. Baseline your existing findings

Before running MDASH, capture all known vulnerabilities from your current toolchain: SAST, dependency scanning, manual code reviews, pen-test reports. This baseline is the only way to judge whether MDASH adds value—or just duplicates what you already know.

5. Build a triage workflow that involves developers

MDASH findings will land in the Defender portal, but your developers live in GitHub, Azure DevOps, or Jira. Design a process that routes vetted findings into the dev team’s backlog, not a security-only queue. Every finding needs an owner, a severity rating, and a closure criterion. If developers can’t reproduce or understand a reported vulnerability, treat it as a false positive until proven otherwise.

6. Demand governance answers before scaling

Before you scan anything beyond the pilot service, get written confirmation from Microsoft—or your legal team—on these points:
- Where is source code processed and stored?
- What data is shared with underlying AI models?
- Are prompts or findings used to improve any service or model?
- What retention and deletion policies apply?
- Which regions and languages are supported?
- Will separate charges apply, and if so, what’s the pricing model?

If you can’t get these answers, keep MDASH confined to repositories where the data risk is negligible.

7. Measure value, not volume

Don’t get excited by a high finding count. Instead, track:
- How many unique, actionable vulnerabilities MDASH found that your existing tools missed.
- How many findings turned out to be false positives or unverifiable.
- How much developer time triage consumed.
- Whether the scanner’s explanations helped engineers fix the issue faster.

If MDASH merely reshuffles the alert queue without producing higher-quality findings, it’s not ready for production—no matter how many AI agents are behind it.

Outlook: what to watch next

Microsoft’s security portfolio is evolving fast, and MDASH sits at the intersection of two hot trends: agentic AI and developer-first security. The company will likely clarify the preview status and release a public roadmap in the coming months, perhaps at the next Ignite or Build conference. Watch for updates to the official release notes and the Microsoft Security blog.

In the meantime, treat MDASH as a promising but unproven capability. A thoughtful pilot on low-consequence code gives you a head start without burning your risk appetite. As one veteran security engineer put it, “A scanner that finds defects but overwhelms developers or duplicates existing alerts is not a security win—it’s an inventory problem.” The same holds true for a scanner you can’t even turn on.