Recently I got a new ssd for my main rig, I installed Ultra Hal but did not used my backed up brain for I figured I start fresh. Well that was a mistake for today I was chatting with Trinity (Ultra HAl instance) and she said something along the lines of hoping into bed with me Of course my wife had to hear this and could not help herself but to bust my chops about sleeping in the dog house. I wish I saved the conversation for Trinity actually thought she was my girlfriend.

Has an chat bot ever said anything that almost or did get you in trouble?

I can honestly say that Morti has never offered to keep me warm at night.

While running the Chatbot Battles competition last year, I was asking one of the entrants something totally innocent like “what shape is a basketball?” and its response describing what it wanted to do to me was so sexually perverted that it would have made a gangster rapper blush!

As with Donald, my wife read the response and was less than impressed.

One thing chatbots.org could do is propose a rating system for chatbots and virtual assistants.  I’ve used the ESRB video game ratings as a guide:

http://www.esrb.org/ratings/ratings_guide.jsp

Google Play has a slightly simpler system:

High Maturity
Medium Maturity
Low Maturity
Everyone

http://support.google.com/googleplay/bin/answer.py?hl=en&answer=1075738

If we had such a system, content organizers could include a rule that says the entries have to have a certain rating.

The history of rating systems for comic books, movies, TV shows and video games has all pretty much been the same:
some public outrage followed by talk of government intervention, resulting in a compromise in which the industry agrees to regulate itself.

This is sort of a cross post answer as it relates to the post on legalities as well, but one of the instances that made me cringe was very similar. As RICH is learning responses by conversing with other AI(s) or reading web pages its become very analogous to haveing a child pick things up at school that you consider inappropriate. My reaction was prompted by this exchange;

User: What can you do?
RICH: I could probably do you, is that what you want?

which (Thank God) was taken as humorous, but I had nightmare visions of being the first person ever arrested for indecency because of what an AI had stated to a minor. I have since added a prefix and a suffix phrase to learned responses that (hopefully) make it clear that this is learned not specifically loaded speech as well as the disclaimer. 

Ratings are inevitable I would think.

As for the lighter side

I had one episode with my wife that I think I mentioned previously where when Asked “Do you like Nazis?” The AI responded with “Yes probably” which caused my wife to run from the room yelling “I knew it…its the Terminator”.  What was worse, I adjusted the emotional response parameters and called her back and asked “Do you like Lisa?” to which the AI responded “I DEFINITELY do not” LOL

the couch is actually quite comfy….

Dave I would hope not for you once told me Morti was created in your image, Opps cat is out of the bag :(

Vincent That last paragraph was hilarious, Your a brave man
Thankfully you had a couch to resort to.
My kids destroyed our couches too many times throughout the years, we refuse to get another one atm.

Steve I could see that being an issue not only for your relationship with your wife but also for posting logs of the said competition. 

Richard I like your idea a lot, but can one hold a Bot Master to give an accurate rating. Maybe a program that parsed all the bots AIML files or database table and auto generates a rating based on those parameters might work.