A hard-coded credential in MuseBot, a bot management platform, means any unauthenticated attacker who knows the default session key can craft a signed cookie claiming to be an admin. According to the NVD advisory for CVE-2026-119660, the key in question is the literal string muse_bot_session_key, baked into versions before 2.0.0. Once you sign a cookie with that key and set a user_id, you have full administrative access. Password hashes, chat records, bot configuration, the ability to kill running bots, all of it.

This is a textbook secrets-management failure, not a subtle one. Hard-coded secrets in session handling are the kind of thing that gets introduced early in a project's life and never rotated because nobody built a rotation process in the first place. The fix exists in 2.0.0, but the architectural lesson is broader: any system that uses a symmetric signing secret for session integrity needs that secret treated the same way you'd treat a private key. Generated at install time, stored in a secrets vault, rotatable without downtime.

The IAM angle here is about what those sessions protect. If MuseBot's admin interface is federated into your SSO or sits behind an identity proxy, an attacker who can forge a session cookie may bypass all of that. The authentication layer your IdP provides is irrelevant if the application trusts a cookie it signed itself using a well-known key. I've seen this pattern show up in internal tooling and third-party integrations more than I'd like to admit. Teams assume the app is "behind the VPN" and stop there. That's not a control, that's a hope.

If MuseBot is anywhere in your environment, the immediate action is clear: upgrade to 2.0.0 now, then check whether any existing sessions issued under the old key need to be invalidated. Don't assume your existing session store is clean just because you patched the binary.

For anything else running a symmetric session secret, pull up your secrets inventory and ask two questions: is this secret generated uniquely per deployment, and when was it last rotated? If you can't answer both, that's your Monday morning audit right there. Generic tooling and internal bots are where hard-coded secrets hide longest because nobody treats them like production systems until something goes wrong.