In abscence of Newspaz:
There was, a while a go, quite a bit of testing done with bots. The bots, however, proved unstable. They could crash servers they entered. They had to have their programming changed for each new level they played in. It took a few months for each level, and it was unreasonably hard. And the final project AI was week, it played like a n00b with rabies. The perfect behavor was near-impossible to program, and thus the project was deemed not worth the effort. Issue 2 of Stuff is Good Magazine is scheduled to have an article on this entitled "The Bots vs. The Programmers".