>they think it'll be more secure (not)

for these types of tasks or LLMs in general?