The Wikipedia article on food tasters explains that human food tasters had multiple functions:
- The safety of the food may be determined by observing whether or not the food taster subsequently becomes ill. However, food tasting is not effective against slow-acting poisons that take a long time to produce visible symptoms.
- The food taster may also prepare and serve food, so they can be even more diligent in preventing someone from poisoning the food.
- In the event the target falls ill or dies, the similar illness or death of the taster provides evidence of deliberate poisoning.
Moreover, using animals would have been much less reliable, particularly when testing for an unknown poison. The most common modern method of testing lethality of poison today is to find the median lethal dose for a variety of animal species. This is known as the LD50 test. As Wikipedia explains:
There can be wide variability between species... what is relatively safe for rats may very well be extremely toxic for humans (cf. paracetamol toxicity), and vice versa. For example, chocolate, comparatively harmless to humans, is known to be toxic to many animals. When used to test venom from venomous creatures, such as snakes, LD50 results may be misleading due to the physiological differences between mice, rats, and humans. Many venomous snakes are specialized predators on mice, and their venom may be adapted specifically to incapacitate mice; and mongooses may be exceptionally resistant. While most mammals have a very similar physiology, LD50 results may or may not have equal bearing upon every mammal species, such as humans, etc.
So while there may be particular poisons for which particular animal species could be used to test somewhat reliably, even having a few different animal species to test a food sample would not be reliable for a full range of possible poisons. And as @SJuan76 points out in the comments, human tasters may also provide useful subjective information about the taste or effect of any potential poison.