ChatNannies Last All Summer Long

This story has been growing in size for the past week, and now it seems definite. It's a hoax. It began with a story in New Scientist last week describing an artificial intelligence program designed by Jim Wightman, an IT consultant from Wolverhampton. This program, called ChatNannies, supposedly scours internet chat rooms pretending to be a child and luring pedophiles into conversation with it. Once it identifies a pedophile, it reports them to the authorities. Sounds great, but if it really does what is claimed of it, it would be the most advanced artificial intelligence software in the world. And created by a guy working out of his home, no less. Almost immediately, people were skeptical. Waxy.org has pretty thoroughly debunked Wightman's claims. There's also some interesting material over at overstated.net where a guy describes his experience chatting online with a 'Nanniebot.' The Nanniebot really does seem eerily human, which is because it almost certainly is human... i.e. Wightman typing away at his keyboard.
I figure this whole thing falls into the hoax genre of 'Amazing Inventions that Can't Be Examined.' It's an old, old modus operandi of hoaxers. Come out with a miraculous new invention, but simultaneously refuse, for one reason or another, to let people inspect it. For examples, you can go all the way back to the Great Chess Automaton of the late 18th century, or Redheffer's Perpetual Motion Machine from the early 19th century. A reporter from the Guardian, Ben Goldacre, is trying to get Wightman to allow him to inspect the ChatNannies program. But so far, he's had no luck.
One unanswered question in all of this is: why did New Scientist ever believe Wightman's claims to begin with? If it weren't for New Scientist publishing the story, it would never have received a fraction of the attention that it already has.

---