Asking whether a healthcare tool or intervention is "safe" almost always really means asking if it is "safe enough." First, almost all healthcare tools and interventions introduce their own inherent risks. Second, almost all tools and interventions can be negligently or deliberately administered wrongly. The Supreme Court of India is considering whether advance directives are safe enough. One key concern is the possibility of coercion or fraud by greedy relatives.
In the United States, there have certainly been cases where advance directives have been abused (see, e.g. Van Note). But the number of these cases is microscopic relative to the millions of advance directives that have helped assure value-congruent care since 1976.
Yes, MAID, POLST, and other tools are not perfect. But perfection is not a reasonable goal. Denying access to imperfect (yet mostly safe) options denies material benefits to multitudes. This is not the standard in healthcare or in the design of cars, toasters, or almost anything else.