Question

1. what limitation of many current a.i. systems does the video highlight as a safety concern?

they never make biased decisions
they are trivial to audit and fully transparent
they always require no data to function
they are complex and often difficult or impossible to fully understand or audit

Explanation:

The question is about identifying a limitation of AI systems. AI systems, especially complex ones like neural networks, are often black - box models. They are complex and difficult to fully understand or audit. They do require data to function (so the option "They always require no data to function" is wrong). They are not trivial to audit (so "They are trivial to audit and fully transparent" is wrong). And they can make biased decisions (but the video in question highlights complexity and auditability as a safety concern, not the "never" part of "They never make biased decisions" which is also factually incorrect as AI can be biased).

Answer:

They are complex and often difficult or impossible to fully understand or audit.