Two perfect 10.0s in Azure, a Chrome Android sandbox escape, and an Entra ID auth bypass
Azure AD B2C and Azure AI Language both score CVSS 10.0 with no auth required. Chrome on Android has a 9.6 WebGL sandbox escape, Entra ID has a 9.1 auth bypass, and Copilot Studio has a 9.3 signature verification failure.
Five critical vulnerabilities dropped and none of them are small. Two are CVSS 10.0: an authorization bypass in Azure AD B2C (CVE-2026-83711) that lets an unauthenticated attacker take over accounts, and a missing-auth bug in Azure AI Language (CVE-2026-70352) that hands over full control with zero credentials. Pair those with an Entra ID auth bypass, a Chrome Android sandbox escape, and a Copilot Studio signature verification failure, and you've got a full plate. None are exploited in the wild yet, but don't wait on these.
Today's CVEs
Sorted by urgencyCVE-2026-62916
NVDAn attacker can bypass authentication in Microsoft Entra ID by using an alternate path, gaining elevated privileges without valid credentials. No user interaction is needed. Because Entra ID is your identity backbone, a successful exploit here could give an attacker control over accounts, roles, and access policies across your tenant.
- Affected estate
- Anyone using Microsoft Entra ID (formerly Azure AD) for identity and access management
- How to check
- Review the Microsoft 365 Service Health dashboard and Entra ID admin center for advisories or required configuration changes tied to this CVE.
- Included because
- unauthenticated; network-exploitable; CVSS 9.1; identity provider is highest-value target
- Action
- Apply any Microsoft-published mitigations, review Conditional Access policies for unexpected changes, and audit sign-in logs for anomalous privilege escalation.
- Why it matters
- An unauthenticated attacker gaining elevated privileges in your identity provider could compromise every service that trusts it.
- Source
- Microsoft Security Response Center
Evidence trail
- NVD: View source
CVE-2026-83711
NVDAn attacker can manipulate a user-controlled key to bypass authorization in Azure AD B2C, escalating privileges over the network with no authentication required. CVSS 10.0, which means no mitigating factors: no user interaction, no special conditions, full impact. If you use Azure AD B2C for customer-facing authentication, an attacker could potentially take over arbitrary accounts or gain admin-level access.
- Affected estate
- Anyone running customer-facing apps that use Azure Active Directory B2C for authentication
- How to check
- Review the Azure portal under Azure AD B2C for advisories. Check Microsoft 365 Service Health and the MSRC advisory page for this CVE.
- Included because
- unauthenticated; CVSS 10.0; network-exploitable; customer identity service
- Action
- Apply Microsoft's fix or mitigation, audit B2C sign-in and audit logs for unauthorized privilege changes, and review custom policy configurations.
- Why it matters
- A CVSS 10.0 unauthenticated privilege escalation in your customer authentication service could expose every customer account.
- Source
- Microsoft Security Response Center
Evidence trail
- NVD: View source
CVE-2026-70352
NVDAzure AI Language is missing authentication on a critical function, letting an unauthenticated attacker escalate privileges over the network. CVSS 10.0. If you're running Azure AI Language services, an attacker could gain full control without needing any credentials at all.
- Affected estate
- Teams using Azure AI Language (formerly Text Analytics, Language Understanding, or related Azure Cognitive Services)
- How to check
- In the Azure portal, review your Cognitive Services or Language resources. Check MSRC for confirmation that the server-side patch has been deployed to your region.
- Included because
- unauthenticated; CVSS 10.0; network-exploitable; no user interaction required
- Action
- Restrict network access to Azure AI Language endpoints using virtual network rules or private endpoints. Confirm Microsoft has applied the server-side fix.
- Why it matters
- Missing authentication on a critical function means zero barrier to exploitation, and a successful attack could give full control over the service.
- Source
- Microsoft Security Response Center
Evidence trail
- NVD: View source
CVE-2026-85050
NVDA crafted HTML page can trigger an out-of-bounds write in Chrome's WebGL rendering on Android, letting an attacker run arbitrary code outside the browser sandbox. That's a full sandbox escape: visiting the wrong page could compromise the device. CVSS 9.6. This only affects Chrome on Android prior to version 152.0.7977.82.
- Affected estate
- Anyone managing Android devices with Google Chrome installed, including MDM and fleet operators
- How to check
- Open Chrome on the device, go to Settings > About Chrome, and confirm the version is 152.0.7977.82 or later. In an MDM, query installed app versions.
- Included because
- sandbox escape; CVSS 9.6; no user interaction beyond visiting a page; widely deployed browser
- Action
- Push Chrome update to 152.0.7977.82+ via the Play Store or your MDM managed app update policy.
- Why it matters
- A sandbox escape from a web page means a single malicious link could fully compromise an Android device.
- Source
- Google Chrome Releases / Chromium Security
Evidence trail
- NVD: View source
CVE-2026-80098
NVDCopilot Studio fails to properly verify cryptographic signatures, letting an unauthenticated attacker escalate privileges over the network. CVSS 9.3. If you build or host bots through Copilot Studio, an attacker could forge or bypass signature checks to gain elevated access without any credentials.
- Affected estate
- Teams building or deploying bots and AI agents through Microsoft Copilot Studio (formerly Power Virtual Agents)
- How to check
- Review the MSRC advisory and Microsoft 365 Service Health for fix deployment status. Check the Copilot Studio admin center for any flagged advisories.
- Included because
- unauthenticated; CVSS 9.3; network-exploitable; cryptographic control bypass
- Action
- Confirm the server-side fix is applied. Audit bot permissions and published agent configurations for unauthorized changes. Review authentication settings on deployed bots.
- Why it matters
- Broken signature verification means an attacker can forge trust, potentially hijacking bot workflows or gaining access to connected data sources.
- Source
- Microsoft Security Response Center
Evidence trail
- NVD: View source
One email, every Wednesday morning.
SubscribeFrom the field notes
From this beat
Read the rest of the field notes →