Three Vendors Patch Critical Flaws Rated 10.0 and 9.5

Three software vendors have released fixes for 11 vulnerabilities spanning their enterprise products, including a credential-reuse bug that earned the maximum severity score of 10.0 and an unauthenticated agent-impersonation flaw rated 9.5. HashiCorp, Veeam, and the Django Software Foundation published security bulletins between July 28 and August 5, 2026, with patched versions now available for Terraform MCP Server, Veeam Service Provider Console, and Django. None of the 11 flaws appear in CISA's Known Exploited Vulnerabilities catalog, and the vendors say they haven't seen active exploitation.

The most severe issue, CVE-2026-16498, sits in HashiCorp's Terraform MCP server and affects deployments running in stateless HTTP mode. The flaw allows one user's Terraform token to be recycled for subsequent users' requests, no matter what credentials they provide. The underlying Model Context Protocol library doesn't create unique session identifiers, and the server's credential cache depended on those identifiers to separate users. The second-highest-rated bug, CVE-2026-58073 with a CVSS 4.0 score of 9.5, lets an unauthenticated attacker impersonate a managed agent in Veeam Service Provider Console and steal that agent's credentials, though its CVSS vector marks attack complexity as high. Veeam's second critical flaw, CVE-2026-58072 rated 9.0, permits arbitrary file writes on the management server that can escalate to remote code execution with a low-privilege account. Django's highest-severity issue, CVE-2026-15307, lives in GeoDjango's spatial lookups and can write files to disk or trigger network requests, potentially leading to remote code execution when a staff user with view permission on a registered model containing a spatial field exploits the path.

The reports describe exposure as configuration-dependent across all three products. HashiCorp's vulnerabilities hit only the Streamable HTTP transport, not the stdio mode used for local single-user setups, and the 10.0-rated stateless bug requires deliberate operator enablement while the 8.9-rated stateful-mode flaw (CVE-2026-16496) affects the default configuration for central deployments. Veeam's four bugs affect Service Provider Console version 9.2.1.33875 and every earlier version 9 build, with the fix landing in build 9.3.0.35057 released July 29. Django's GeoDjango flaw requires a staff account with view permission, so reconnaissance that identifies PostGIS-backed Django installations "stops a step short of an actual way in," according to the advisory. The report notes a discrepancy in HashiCorp's published version ranges: the umbrella bulletin lists affected versions 0.2.1 through 1.0.0, while individual CVE records start at 0.3.0, though both agree version 1.1.0 is the first fixed release.

Operators should upgrade Terraform MCP Server to version 1.1.0 or later, Veeam Service Provider Console to 9.3.0.35057, and Django to 6.0.8 or 5.2.17, the advisories state. For HashiCorp users who can't upgrade immediately, the report recommends restricting network access to the Streamable HTTP listener to trusted users and treating MCP session identifiers as sensitive values. The Veeam bulletin marks this as the second critical patch cycle for the console in roughly three months, following a May fix for CVE-2026-32998, a 9.4-rated remote code execution bug tied to alarm script execution. Django's fix for CVE-2026-15307 blocks dict values and strings that aren't valid GEOSGeometry values in spatial lookups, a backward-incompatible change, while direct model-field assignments still accept those types. The report highlights that Django's GIS code already attracted attacker attention in February when CrowdSec observed exploitation of CVE-2026-1207, a SQL injection flaw, and released detection rules after seeing steady probing aimed at locating PostGIS-backed Django deployments. Patching windows for enterprise software increasingly compress as threat actors refine their reconnaissance capabilities, and organizations running multi-tenant infrastructure face heightened risk when credential-isolation assumptions break down at the library level.