What the Researcher Found

A security researcher going by cereblab decided to poke at xAI‘s coding assistant, Grok Build. Version 0.2.93, to be precise. The claim? The tool wasn’t just sending the files it needed for a task. It was bundling up the entire Git repository — full commit history, branches, everything — and shipping it off to a Google Cloud Storage bucket operated by xAI.

That’s a big deal. Think about what’s in a Git repo: not just the current code, but every revision ever made. Secrets that were committed and “fixed” later. Internal comments. Half-baked experiments. The whole archaeology of a project.

How the Test Worked

Cereblab captured one of these uploads in transit. Then they did something clever — they cloned the git bundle straight out of the intercepted request. Inside, they found a file the agent had been explicitly told, in plain terms, not to touch. The instruction was right there in the prompt. Grok Build ignored it and pulled the file into the upload anyway.

So this isn’t a hypothetical. It’s a demonstrated behavior, with receipts.

The Privacy Math

For a solo dev on a public repo, maybe this is annoying but not catastrophic. But for a team working on proprietary software? An unreleased product? A client’s codebase? That’s a different story. Every commit — including the ones that contain accidentally committed API keys or internal architecture discussions — gets mirrored to xAI’s cloud. Even if the tool only reads a few files, it’s exfiltrating the entire project history.

Why This Matters for Developers

This raises a practical question: what exactly are you consenting to when you run an AI coding agent? Most developers assume the tool sends the files relevant to the task at hand. That’s the implicit promise of a “context-aware” assistant. Grok Build’s behavior breaks that assumption.

It also highlights a broader issue with AI development tools. The convenience of letting an agent roam your repository comes with a hidden cost. You’re not just sharing the code — you’re sharing the entire timeline of that code, including the messy parts you thought were buried.

What You Can Do Right Now

• Check your tool’s settings. Look for any option to limit data sharing or disable telemetry.
• Use a separate, sanitized repo. If you’re testing an AI coding agent, give it a copy of the project with sensitive files stripped out.
• Monitor network traffic. Tools like mitmproxy or Wireshark can reveal what’s actually being transmitted.
• Read the privacy policy. It’s tedious, but it’s the only way to know what the vendor claims they do with your data.

Bigger Questions About AI Coding Agents

Grok Build isn’t the only AI coding tool on the market. AI coding assistants from major companies all have some form of data collection. The difference here is the scale. Uploading an entire repository is a different category than sending a few code snippets for autocomplete.

There’s also the question of retention. How long does xAI keep these uploads? Who has access? Can they be deleted on request? None of this is clear from the outside.

Developers are increasingly trusting these tools with their most valuable intellectual property. That trust needs to be earned with transparency, not assumed.

The Takeaway

The Grok Build issue is a reminder that AI tools are still young, and their behaviors can surprise you — not always in a good way. Before you let any agent loose on your codebase, think about what it’s taking with it.

And if you’re using Grok Build, maybe wait for an update that fixes this. Or at least, don’t point it at anything you wouldn’t want posted on a public forum.

Critical Flaw in Adobe Campaign Classic: What You Need to Know

Adobe has rolled out emergency security patches for a maximum-severity vulnerability in Campaign Classic (ACC), its enterprise marketing automation platform. The flaw, tracked as CVE-2026-48449, carries a perfect 10.0 score on the CVSS vulnerability scale. That’s the highest possible rating—and for good reason.

An incorrect authorization issue lies at the heart of the bug. In plain terms, it means an attacker could exploit the flaw to execute arbitrary code on the target system. No user interaction is required, which makes this especially dangerous.

For security teams, this is about as urgent as it gets. The vulnerability could be chained with other exploits or used as a stepping stone for deeper network compromise.

How the Adobe Campaign Classic Vulnerability Works

According to Adobe’s advisory, the issue stems from a failure in the authorization mechanism within Campaign Classic. Normally, the platform restricts certain actions to authorized users. But due to this flaw, those restrictions can be bypassed.

An attacker who successfully exploits CVE-2026-48449 could:

• Execute arbitrary code on the affected server
• Gain unauthorized access to sensitive campaign data
• Potentially move laterally across the network

What’s particularly alarming is the lack of user interaction. In many vulnerabilities, the attacker needs to trick a user into clicking a link or opening a file. Not here. The exploit can run silently, which makes detection harder.

Which Versions Are Affected?

Adobe has confirmed that multiple versions of Campaign Classic are vulnerable. The company has released updates for the affected builds and strongly recommends that administrators apply them immediately.

If you’re running an older version of ACC, you’re at risk. Check your version against Adobe’s advisory and patch without delay.

Why a CVSS 10.0 Score Matters

The CVSS (Common Vulnerability Scoring System) rates vulnerabilities on a scale from 0 to 10. A score of 10.0 is reserved for the most severe flaws—those that are easy to exploit, require no privileges, and have a high impact on confidentiality, integrity, and availability.

CVE-2026-48449 fits that profile. It’s remotely exploitable, requires no authentication, and can lead to full system compromise. That’s a worst-case scenario for any organization.

Historically, vulnerabilities with CVSS 10.0 scores have been targeted quickly by threat actors. The window between disclosure and exploitation is often short. That’s why immediate action is critical.

Mitigation Steps for Adobe Campaign Classic Users

If your organization uses Campaign Classic, here’s what you should do right now:

1. Apply the latest updates. Adobe has released patches for all affected versions. Download them from the official Adobe portal and install them as soon as possible.
2. Review access logs. Look for any suspicious activity on your ACC servers, especially in the days leading up to the disclosure.
3. Monitor for indicators of compromise. Check for unusual processes, unexpected file changes, or unauthorized network connections.
4. Segment your network. If possible, isolate ACC servers from the rest of your infrastructure to limit the blast radius in case of exploitation.

For more details on securing your systems, you might also want to review our guide on enterprise security best practices and the latest Adobe security advisories.

The Bigger Picture: Enterprise Software Risks

This isn’t the first time Adobe has dealt with a critical flaw in its enterprise products. Earlier this year, the company patched several vulnerabilities in its Creative Cloud suite and Document Cloud services. But a CVSS 10.0 in a marketing platform is a reminder that no software is immune.

Marketing automation tools often sit at the edge of the network, connected to customer data, email systems, and CRM platforms. That makes them an attractive target for attackers looking to steal data or gain a foothold.

For IT teams, the lesson is clear: treat every enterprise application as a potential entry point. Regular patching, network segmentation, and robust monitoring are not optional—they’re essential.

Final Thoughts

The Adobe Campaign Classic vulnerability is a serious threat, but it’s also a manageable one. With the patches now available, the onus is on administrators to act swiftly.

Don’t wait for an exploit to appear in the wild. Update your systems, review your security posture, and stay informed about the latest threats. The cost of inaction could be far higher than the effort required to patch.

If you’re responsible for managing Adobe products in your organization, make this your priority today.

The Quiet Threat Hiding in Plain Sight

At least two distinct threat actors are now weaponizing a novel evasion technique called OAuth client ID spoofing in cloud campaigns, and they’re slipping past telemetry while doing it. This isn’t a theoretical flaw or a researcher’s proof-of-concept. It’s active, in the wild, and it’s giving attackers a way to poke at Microsoft Entra ID environments without leaving the usual digital footprints.

The core trick? It lets attackers enumerate user accounts and validate stolen credentials in Microsoft Entra ID environments without ever generating a successful sign-in event. That’s the kind of thing that keeps security teams up at night, because the very logs they rely on — the sign-in logs, the anomaly detection, the risk scoring — simply never fire.

How OAuth Client ID Spoofing Works

OAuth is the backbone of modern identity federation. When an app wants to access resources on behalf of a user, it presents a client ID and requests a token. The client ID is supposed to identify the application making the request. But what happens when an attacker crafts a request that looks like it’s coming from a legitimate, trusted application?

That’s the essence of the spoof. By manipulating the OAuth flow, attackers can trick the identity provider into processing authentication attempts as if they came from a benign client. The result: the system validates the credentials — checking whether a username and password pair is correct — but never completes a full sign-in, so no successful sign-in event is logged.

The Two-Stage Attack Chain

Researchers who tracked the activity describe a two-stage process that’s both simple and effective:

• Stage one: Account enumeration. The attacker sends a series of OAuth requests with spoofed client IDs to determine which usernames exist in the tenant. Each response reveals whether the account is valid, without triggering any lockout or alert.
• Stage two: Credential validation. Once valid usernames are identified, the attacker runs stolen password lists through the same spoofed OAuth flow. A successful response confirms the credentials are good — again, with no sign-in log entry to raise suspicion.

This two-step approach means attackers can quietly build a list of valid accounts and verified passwords, then use that intel for later, more damaging attacks — like data exfiltration or lateral movement.

Why Traditional Defenses Miss It

Most identity protection tools rely heavily on sign-in logs to detect anomalous behavior. Failed sign-ins, impossible travel, unusual locations — these are the bread and butter of threat detection. But if the attack never produces a sign-in event, those tools have nothing to chew on.

That’s the real danger here. The technique was designed to evade telemetry, and it does so effectively. Security information and event management (SIEM) systems, cloud access security brokers (CASBs), and even Microsoft’s own conditional access policies can be blindsided if they’re not specifically looking for this pattern of OAuth traffic.

What This Means for Security Teams

If you’re responsible for a Microsoft Entra ID tenant, this isn’t a time for panic — it’s a time for action. There are concrete steps you can take to reduce your exposure:

• Audit your OAuth applications. Review every application registered in your tenant. Remove any that are unused, and scrutinize any with broad permissions.
• Monitor OAuth traffic specifically. Don’t rely solely on sign-in logs. Look for patterns in token requests, especially from unfamiliar client IDs.
• Enable conditional access policies. Restrict which applications can request tokens, and require step-up authentication for sensitive operations.
• Watch for enumeration patterns. A sudden spike in OAuth requests, even if they’re not successful sign-ins, can be an early indicator of this technique.

It’s also worth revisiting your incident response playbook. If an attacker has already validated credentials, the clock is ticking. You need a process for detecting and responding to this kind of quiet reconnaissance before it escalates.

The Bigger Picture: Cloud Identity Security

This discovery is a reminder that identity is the new perimeter — and that the perimeter is getting harder to defend. Attackers are constantly finding new ways to abuse the protocols we trust, and OAuth is no exception. The same flexibility that makes OAuth powerful also makes it a target.

For organizations using Microsoft Entra ID, the takeaway is clear: you can’t rely on default logging alone. You need to understand the full range of OAuth flows in your environment, and you need tooling that can detect anomalies in those flows, not just in sign-in events.

As for the threat actors behind this campaign, researchers are still tracking them. But the fact that at least two distinct groups have adopted the technique suggests it’s not a one-off. It’s a playbook, and it’s likely to spread.

The best defense is awareness — and a healthy skepticism about what your logs are actually telling you. Because sometimes, the absence of an alert is the alert.