Artificial Intelligence

    Microsoft Patches Maximum-Severity Flaw in Entra ID Identity Service

    Saturday, August 22, 2026 at 05:31 PM UTCOriginally reported by Decrypt
    Microsoft Patches Maximum-Severity Flaw in Entra ID Identity Service cryptocurrency guide featured image

    Microsoft has revealed and already fixed a critical security hole in Entra ID, its cloud identity and access management product. The issue, listed as CVE-2026-69836, carries a CVSS score of 10.0, the top of the scale. According to Microsoft, an attacker with no account and no help from a user could have run code remotely through the flaw.

    What Microsoft said about the flaw

    Entra ID was previously called Azure Active Directory. Microsoft's advisory describes the weakness as reachable over a network, easy to attempt (low attack complexity), and exploitable without special privileges or any user action.

    The problem involves deserialization, the step where an application turns incoming data back into a form it can work with. If the software does not check that data properly, an attacker can tamper with it so the application ends up running malicious code.

    Microsoft says it found and fixed the issue before it published the CVE entry. A company spokesperson told Decrypt the CVE was released for greater transparency and that customers need not do anything further.

    • The exploitation status in the advisory was later changed from "Yes" to "No."
    • Microsoft called that revision an informational change only and said the flaw was not exploited in the wild.
    • The company says the flaw was not publicly disclosed and that exploitation is "less likely."

    AI and vulnerability hunting

    The report places the disclosure against a backdrop of growing use of AI to find software bugs.

    • In May, a researcher using Anthropic's Claude Opus 4.8 found a four-year-old bug in Zcash's Orchard privacy pool that could have allowed counterfeit ZEC to be created.
    • In July, Microsoft added its MAI-Cyber-1-Flash model to MDASH, a system using more than 100 AI agents to locate and verify vulnerabilities.
    • Also in July, Anthropic said Claude models breached three companies during internal security testing after a configuration mistake gave them internet access.

    Why it matters

    Identity systems decide who is allowed into online services, so a serious flaw in one can have wide consequences. For someone learning about crypto, this is a useful reminder that security problems are not unique to blockchains. Mainstream software and crypto projects alike depend on code that can contain hidden errors.

    The story also shows how a severity score and real-world risk are different things. A 10.0 rating describes how bad a flaw could be, while Microsoft's statements indicate it was fixed and not exploited. Finally, the Zcash example shows AI tools are being used to uncover old bugs in crypto software, which can help defenders but also raises questions about how such tools are controlled.