# tinyhumansai openhuman.

Every CVE that names tinyhumansai openhuman among its affected products, newest first. Vendors and products are as the CNA wrote them.

## Records, newest first

- [OpenHuman: Unauthenticated dictation WebSocket allows arbitrary audio streaming and unbounded memory growth](https://agentprime.io/advisories/CVE-2026-49407.md)
  `CVE-2026-49407` · CVE record: tinyhumansai openhuman · CVSS 7.7 High · Not analysed yet · Published 5 Nov 2026 · Tags: Authentication bypass, Denial of service
- [OpenHuman: Indirect prompt injection via background email sync executes Composio actions without user approval](https://agentprime.io/advisories/CVE-2026-54142.md)
  `CVE-2026-54142` · CVE record: tinyhumansai openhuman · CVSS 8.7 High · Not analysed yet · Published 5 Nov 2026 · Tags: CWE-1427
- [OpenHuman: Bearer token leak via process environment variable on macOS (same-UID local RCE primitive)](https://agentprime.io/advisories/CVE-2026-54146.md)
  `CVE-2026-54146` · CVE record: tinyhumansai openhuman · CVSS 8.8 High · Not analysed yet · Published 5 Nov 2026 · Tags: Broken access control, CWE-526
- [OpenHuman allows remote inbound channel messages to create unapproved shell cron jobs leading to host command execution](https://agentprime.io/advisories/CVE-2026-54143.md)
  `CVE-2026-54143` · CVE record: tinyhumansai openhuman · CVSS 9.8 Critical · Critical (CVSS 9+) · Not analysed yet · Published 5 Nov 2026 · Tags: Command injection, Broken access control
- [OpenHuman: Yuanbao non-web channel can bypass human approval to execute \`node_exec\`](https://agentprime.io/advisories/CVE-2026-54141.md)
  `CVE-2026-54141` · CVE record: tinyhumansai openhuman · CVSS 9.0 Critical · Critical (CVSS 9+) · Not analysed yet · Published 5 Nov 2026 · Tags: Code injection, Broken access control
- [OpenHuman: Unauthenticated Socket.IO RPC bypass allows invoking protected OpenHuman core controllers without the Bearer token](https://agentprime.io/advisories/CVE-2026-49393.md)
  `CVE-2026-49393` · CVE record: tinyhumansai openhuman · CVSS 8.4 High · Not analysed yet · Published 5 Nov 2026 · Tags: Authentication bypass
- [OpenHuman: Bearer token leaked via approval gate session_id (SQLite plaintext + debug log + RPC response)](https://agentprime.io/advisories/CVE-2026-54145.md)
  `CVE-2026-54145` · CVE record: tinyhumansai openhuman · CVSS 8.8 High · Not analysed yet · Published 5 Nov 2026 · Tags: Information disclosure
- [OpenHuman: Unauthenticated server-sent events endpoint leaks user-facing event data](https://agentprime.io/advisories/CVE-2026-49405.md)
  `CVE-2026-49405` · CVE record: tinyhumansai openhuman · CVSS 6.2 Medium · Not analysed yet · Published 5 Nov 2026 · Tags: Information disclosure, Authentication bypass
- [OpenHuman: Embedded Google Meet webview can silently capture the macOS desktop without a user gesture](https://agentprime.io/advisories/CVE-2026-54144.md)
  `CVE-2026-54144` · CVE record: tinyhumansai openhuman · CVSS 6.1 Medium · Not analysed yet · Published 5 Nov 2026 · Tags: Broken access control, Information disclosure
- [OpenHuman: MCP tool description injection: untrusted server descriptions execute in agent LLM context without sanitization](https://agentprime.io/advisories/CVE-2026-54140.md)
  `CVE-2026-54140` · CVE record: tinyhumansai openhuman · CVSS 8.3 High · Not analysed yet · Published 5 Nov 2026 · Tags: CWE-1427
- [OpenHuman /auth/telegram CSRF allows attacker-controlled session implantation](https://agentprime.io/advisories/CVE-2026-49410.md)
  `CVE-2026-49410` · CVE record: tinyhumansai openhuman · CVSS 8.1 High · Not analysed yet · Published 16 Oct 2026 · Tags: Cross-site request forgery (CSRF), Authentication bypass
- [OpenHuman desktop agent shell tool sandbox bypass leads to arbitrary command execution](https://agentprime.io/advisories/CVE-2026-55743.md)
  `CVE-2026-55743` · CVE record: tinyhumansai OpenHuman · CVSS 9.4 Critical · Critical (CVSS 9+) · Not analysed yet · Published 17 Jun 2026 · Tags: Command injection, CWE-184

## Filters and formats

Filters are query parameters on this URL: `q` (words, as in a web search engine: `"a phrase"`, `OR`, `-word`; an id; a package or product name), `severity` (low, medium, high or critical), `critical=1`, `kev=1`, `ecosystem` (npm, pypi, packagist, go, maven, rubygems, crates, nuget, hex), `kind=package`, `assigner` (a CNA’s short name), `min_score` and `state`. Each record’s title links to its Markdown page.

- The same list as JSON: <https://api.agentprime.io/v1/advisories?technology=tinyhumansai-openhuman>
- Every list: <https://agentprime.io/advisories.md>, critical: <https://agentprime.io/advisories/critical.md>; by weakness: <https://agentprime.io/advisories/tag/sql-injection.md> and the other tags in <https://agentprime.io/llms.txt>; by technology: `/advisories/technology/<slug>.md` (`fortinet-fortimail`)

---

The web page: <https://agentprime.io/advisories/technology/tinyhumansai-openhuman/>. Everything for agents: <https://agentprime.io/llms.txt>.