LIVE TAPE
OpenClaw 88,412 stars·CVE-2026-25898 disclosed (HIGH, Hermes)·Hermes Agent v2026.4.7 published·Hermes Agent +182 stars (last hour)·OpenClaw v2026.4.6 — credential vault hardening·CVE-2026-26133 patched (NanoClaw)·Pi 5 16GB rumoured for Q3 — recheck guidance·Nanobot +47 stars (last hour)·ZeroClaw v0.4.2 — Apple container fixes·Mac Mini M4 wins quarterly hardware survey·OpenClaw 88,412 stars·CVE-2026-25898 disclosed (HIGH, Hermes)·Hermes Agent v2026.4.7 published·Hermes Agent +182 stars (last hour)·OpenClaw v2026.4.6 — credential vault hardening·CVE-2026-26133 patched (NanoClaw)·Pi 5 16GB rumoured for Q3 — recheck guidance·Nanobot +47 stars (last hour)·ZeroClaw v0.4.2 — Apple container fixes·Mac Mini M4 wins quarterly hardware survey·
PocketClawvol. 1 · 2026

Self-hosted AI security

Threat models, sandbox practice, CVE history and incident-response playbooks for self-hosted AI agents in 2026.

When OpenClaw shipped CVE-2026-25253 in January, ~1,000 public installs were popped before the patch landed. Self-hosted AI security stopped being theoretical that month. This hub collects what we know about it.

The post-OpenClaw-crisis security baseline for any agent you self-host in 2026 looks like this: sandbox-on by default, authenticated dashboard, encrypted credential vault, explicit network egress allowlist, hash-chained audit log, working CVE disclosure channel, and a 48-hour patch SLA on critical issues.

Hermes Agent post-2026.4 ships all of those. OpenClaw 2026.4+ ships most of them. Nanobot ships none of them by design — it's a single-user codebase audited by reading. IronClaw goes further still with gVisor and SAML.

If you're not at this baseline, you're closer to the pre-2026 dangerous defaults than you think. The articles below cover the threat model, the audit checklist and the incident response playbook. The CVE archive lists every disclosed vulnerability we track. The agents and providers pages note who passes and who fails our security baseline.

Self-hosted AI security is not exotic. It's regular hygiene applied to a moderately new attack surface. The mistakes that matter are mistakes we already know how to avoid in other software.

Guides

Agents on this topic

  • OpenClawThe original viral self-hosted AI agent. Post-crisis 2026.4 line is genuinely safer; pre-2026.3 is genuinely dangerous.
  • Hermes AgentPost-OpenClaw safe default. Docker-sandboxed by default, multi-LLM, opinionated. The agent we'd hand a colleague today.
  • IronClawEnterprise zero-trust agent. gVisor + audit logs + RBAC + SAML. $750/seat/year.
  • ZeroClawPrivacy-first. Local LLMs only. Network egress denied at iptables. AGPL-3.0.

Vulnerabilities

  • CVE-2026-25253OpenClaw — 1-click RCE via WebSocket origin bypass
  • CVE-2026-25103OpenClaw — credential storage in plaintext on disk
  • CVE-2026-25898OpenClaw — credential leak via verbose error logs
  • CVE-2026-23912MCP protocol — tool description spoofing across implementations

Terms

Sandbox · Sandboxed tool execution · Approval flow · Prompt injection · RCE (Remote Code Execution) · WebSocket origin validation · CVE-2026-25253