CVE WatchSep 18, 2026

CVE-2026-92702

Cocos AI is a confidential computing system for running AI workloads inside trusted execution environments. In versions up to and including

CVSS

9.1

Critical

EPSS

—

KEV

—

Exploit Today

—

0-100

Published: Sep 18, 2026 · Last modified: Sep 18, 2026 · CWE-346

EPSS · 30d

Not enough EPSS history yet.

Technical description

Cocos AI is a confidential computing system for running AI workloads inside trusted execution environments. In versions up to and including 0.8.2, the intra-handshake attested TLS (aTLS) AMD SEV-SNP verification path does not enforce attestation freshness when the expected reportData value is nil, empty, or omitted, leaving the SEV-SNP policy ReportData unset so the verifier accepts unrelated or stale Evidence not bound to the current connection. A relying party that uses this path without an expected reportData as a trust or authorization decision can be induced to trust an unintended attestation context; a supplied non-empty reportData is still validated. The issue is fixed in version 0.9.0.

Official references

Related CVEs

CVECVSSEPSSKEVRExploitTitleMod.

CVE-2026-927019.1 CRI

—

———trusted execution environments. In versions up to and including 0.8.2, the intra-handshake attested TLS (aTLS) Intel TDX verification path does not copy the expected current-session freshness value into the TDX quote-body policy before quote validation, so structurally valid TDX QuoteV4 Evidence is accepted without checking that its REPORT_DATA field matches the reportData expected for the current session. A relying party using this path can therefore accept Evidence with a mismatched or reused reportData and release application data after the handshake, enabling session-misbinding to an unintended attestation context. The issue is fixed in version 0.9.0.9h CVE-2026-77339—

—

———Process Compose is a scheduler and orchestrator for non-containerized applications. Prior to 1.120.0, the MCP SSE listener in src/mcp/server.go accepts browser-origin requests to /sse and the returned message endpoint without validating the Host header, validating the Origin header, or authenticating the caller. When MCP SSE is enabled, a malicious website can use DNS rebinding to reach the loopback listener and issue MCP requests. If expose_control_tools is enabled, the attacker can enumerate process state, read or search logs, truncate logs, and start, stop, restart, or scale local processes; configured user-defined tools can expose additional commands and output. The Gin REST API token middleware does not protect this separately started MCP listener. This issue is fixed in version 1.120.0.12h CVE-2026-122843.7 LOW

3.1%

——1Mattermost Desktop App versions <=6.2 6.2.2.0 fails to validate the IPC sender in the leaveCall handler which allows a malicious or compromised Mattermost server (or a user with script access to a connected server view) to disconnect an active call belonging to a different connected server via the desktopAPI.leaveCall IPC message. Mattermost Advisory ID: MMSA-2026-0069916h CVE-2026-750254.7 MED

4.3%

——1Mattermost Desktop App versions <=6.2 6.2.2.0 Fixed an issue where Mattermost Desktop did not sufficiently restrict server-rendered content from accessing local or private network resources. Thanks to game0v3r for contributing to this improvement under the Mattermost responsible disclosure policy. Mattermost Advisory ID: MMSA-2026-006982d CVE-2026-771195.9 MED

8.7%

——3A validly signed NSEC3 from an unrelated sibling zone may be accepted as an insecurity proof, downgrading a secure delegation and letting a forged unsigned answer through. This issue affects BIND 9 versions 9.11.0 through 9.18.50, 9.20.0 through 9.20.27, 9.21.0 through 9.21.25, 9.11.3-S1 through 9.18.50-S1, and 9.20.9-S1 through 9.20.27-S1.2d CVE-2026-923606.3 MED

7.5%

——2A weakness has been identified in ag-ui-protocol ag-ui 1.0. The impacted element is the function prepareRunAgentInput of the file agent/agent.ts of the component Event Application Layer. This manipulation of the argument TEXT_MESSAGE_START causes origin validation error. Remote exploitation of the attack is possible. The pull request to fix this issue awaits acceptance.2d