Microsoft SQL Server CVE-2026-65674

Exploitation requires the attacker to hold valid low-privilege authenticated access to the target SQL Server instance (CVSS PR:L) and network reachability to it (AV:N); no user interaction is needed (UI:N). The attacker must be able to supply untrusted data to a SQL Server component that deserializes it (CWE-502). The primary limiting factor is the authentication requirement - a credential-less external attacker cannot exploit this directly - so realistic risk concentrates on instances exposed to untrusted or lower-trust database users (e.g. multi-tenant apps, service accounts, or internal users). Affected builds span SQL Server 2019, 2022, and 2025 (GDR and CU branches per EUVD); instances patched to the listed fixed builds (e.g. 16.0.4305.5, 15.0.4500.2, 17.0.1145.4) are not vulnerable.

This is a serious but conditional risk. The CVSS 3.1 vector (AV:N/AC:L/PR:L/UI:N) scores 8.8 (High) and indicates a network-reachable, low-complexity flaw that nonetheless requires the attacker to already hold valid low-privilege credentials (PR:L) on the SQL Server instance. CWE-502 (deserialization of untrusted data) combined with C:H/I:H/A:H reflects full code execution impact, which aligns with CISA SSVC rating Technical Impact: total. However, the exploitation signals temper the urgency: SSVC lists Exploitation: none and Automatable: no, the CVE is not in CISA KEV, and no public POC is referenced - so there is no public exploit identified at time of analysis. The combination of total technical impact but 'none' exploitation and 'no' automatability suggests a high-impact flaw that is not yet being exploited and is not trivially wormable, making this a patch-priority item (especially where untrusted users have database access) rather than an emergency drop-everything scenario. Key limiting factor to verify: the required privilege level (PR:L) means a fully external, credential-less attacker cannot exploit this directly.

An attacker who already holds low-privilege authenticated access to a SQL Server instance (for example, a compromised application service account or a malicious low-rights database user) sends a crafted serialized object over the network to a component that unsafely deserializes it. The resulting deserialization executes attacker-controlled code with the privileges of the SQL Server process, yielding full confidentiality, integrity, and availability compromise. No public exploit code is identified at time of analysis, and no user interaction is required (UI:N).