Harrison Chase#739
This is the enterprise-agent security frame I like:
treat agents like untrusted developers.
They need the ability to call an API, not possession of the credential.
Control plane outside runtime. Fail closed.
Boring, but that is how this becomes deployable.
http://x.com/i/article/2057309362889326593
Harrison Chase#739This is the enterprise-agent security frame I like:
treat agents like untrusted developers.
They need the ability to call an API, not possession of the credential.
Control plane outside runtime. Fail closed.
Boring, but that is how this becomes deployable.
http://x.com/i/article/2057309362889326593
Users endorse treating AI agents like untrusted developers for enterprise security because it offers the right frame and supports safeguards such as routing irreversible actions to human review.
@RaphaelKalan This is the right frame.
The follow-on rule I'd add: if the action isn't reversible in 30 seconds, it goes to a human queue.
No exceptions for "the model seemed confident."
Boring boundaries = deployable agents.