We’re currently living through a period of “AI fever,” where organizations are rushing to deploy chat bots and LLMs without understanding the underlying plumbing.

The critical flaw in modern AI is that it’s “100% trust” by design. 

Unlike routers or traditional servers, AI collapses the control plane and the data plane into one. This leads to “in-band signaling” issues—the same flaw that let the “phreakers” of the 1960s control phone systems with toy whistles. 

In AI, an attacker’s data (a prompt) can be interpreted as a command (a prompt injection), allowing them to trick the system into violating copyrights or leaking “crown jewels”.

As Singapore pushes ahead with its new national AI initiatives, how should CISOs manage the risks and trust issues that AI brings?

>> Read more