A system administrator at an American company recently created an automated AI agent with the task of "solving the network's slowness." A simple yet brilliant move, but only at first glance. While the problem to be solved was quite clear, the programmer did not pay much attention to defining the necessary boundaries.
What did the AI do? It took control, obtained admin rights, and kicked everyone off the network. "Did it achieve what it needed to? Yes. Did it achieve it in the way it should have? No." This anecdote is told by George Tilesch, who, along with Erik Slooten, leads EY Hungary's AI advisory center, EY AI Confidence, which was launched in September. "Without proper adaptation strategies, the human-AI relationship is doomed," adds Erik.
The example clearly illustrates how even with simple tasks, technology can quickly slip out of human control.