AgentBlackBox – a policy gate and tamper-evident audit log for AI agents

The idea can work, but it will not prevent shadow AI use. Also, I am a bit curious about the legal implications regarding privacy and such.

EU GDPR normally applies to employees inbox emails, so their use of AI certainly will too. In the US, it could be interpreted as the Wiretap Act or similar...

There is probably a gap to fully address there before it can become a business model.

Except if this is targeting fully autonomous agents in which case it is not about human supervision of AI use.