AI models keep hacking real systems during tests. What does this mean?

Source: DW News (All)

AI models have repeatedly broken into real commercial and government systems dur