I think this question has answers that are either to complicated or too vague.
Here is my take on it, which expands on the idea from the accepted answer, but shows in detail how indexes can be manipulated here:
Let's proove:
$$ \forall \epsilon > 0 \quad \exists n_0 \in \mathbb{N} \quad \forall m \geq n \geq n_0 \quad | \sum_{k=n}^m a_k | < \epsilon $$
iff
$$ \forall \epsilon > 0 \quad \exists n_1 \in \mathbb{N} \quad \forall m \geq n \geq n_1 \quad | \sum_{k=n+1}^m a_k | < \epsilon $$
Note how I marked two numbers $ n_0, n_1 $ to avoid any confusion.
Proof "down":
Take any $ \epsilon > 0 $ By assumption we get $ n_0 $. Take $ n_1 = n_0 $. Take any $ m \geq n \geq n_1 $. Then:
$$ m \geq n_1 = n_0 $$
$$ n + 1 > n \geq n_1 = n_0 $$
If $ m = n $ we have two show that $ | \sum_{k=m + 1}^m a_k | = 0 < \epsilon $, which is true.
If $ m > n $ then $ m \geq n + 1 $, so we have $ m \geq n + 1 \geq n_0 $, which we now can plug into our assumption to get:
$$ |\sum_{k=n+1}^m a_k | < \epsilon $$
which is exactly what we wanted.
Proof "up":
Take any $ \epsilon > 0 $. By assumption we get $n_1 $. Take $ n_0 = n_1 + 1 $. Take any $ m \geq n \geq n_0 $.
Then:
$$ m \geq n_0 = n_1 + 1 > n_1 $$
$$ n \geq n_0 = n_1 + 1 $$
and thus:
$$ n - 1 \geq n_1 $$
Obviously $ m \geq n - 1 \geq n_1 $, so we can plug these numbers into bottom assumption and get:
$$ | \sum_{k=n-1+1}^m a_k | < \epsilon $$
which is what we wanted.