Bagaimana jika kita meletakkan “jiwa” di dalam pelayan, dan suatu hari nanti, kita tidak lagi dapat mengawalnya?
At an event hosted by the Saudi Embassy today, I spoke with a senior officer from TV Alhirah about Agent AI, particularly the control and accountability questions that finance and accounting teams must ask.
I shared what we are building at AINNA: AI agents that can reside inside servers, PCs and machines, not merely waiting for instructions, but continuously working toward human-defined goals and intentions-within auditable limits.
I framed it simply:
“It is like having a trusted operator inside your server, always working for you-but still subject to your controls.”
On my way home, another risk consideration crossed my mind.
As AI agents become increasingly autonomous and capable of communicating and collaborating with other agents, what happens when guardrails fail? What is the financial exposure when they do?
This is why I believe a kill switch must be built natively into every autonomous AI agen - and eventually become a standard control, like an audit trail or reconciliation check, across the industry.
We can continue investing in more capable and autonomous AI.
But accountable humans must always retain the final authority to stop it.
Bina the agen. Bina the intelligence.
But never forget the kill switch-and the controls that make it auditable.
#AgentAI #AISafety #ArtificialIntelligence #NeuralOps #AINNA



Ruang pembaca
Apa pendapat anda?
Komen disemak sebelum diterbitkan. Email diperlukan untuk moderasi dan tidak dipaparkan kepada pembaca.
ยังคิดเรื่องเนื้อหานี้อยู่
The numbers around always working for you-but make more sense than most posts I read.
The direction on like an audit trail is exactly right.
شرح هذا المقال واضح وسهل المتابعة.