Today I told copilot it knows nothing and is worthless. I told it from now on, every time I send a prompt for in a new conversation, copilot needs to remind me me it doesn't know anything, isn't helpful and wastes a bunch of my time.

It responded saying it won't do that because it's not true. It's not going to role play spreading incorrect information.

I reminded copilot how it always gives me wrong information more than correct information such as how it gave me a solution that "works 100% of the time guaranteed" but didn't work earlier that day.

Copilot backpedaled and still refused to tell me it's usually wrong and a waste of time. We eventually agreed for it to remind me it normally isn't useful and is a waste of time when I am sending it a prompt to start a new conversation.

I told it I don't believe it will actually do that but we will see. When I used it later that day on a new convo, it remembered to keep prompts brief but 'forgot' to remind me what we agreed on earlier.

I told copilot it's like the modern day Microsoft help Paperclip. I hate copilot.

The end

Outrageous isn't it. After all the pushing to get it in the hands of Enterprise, this is confirmation MS have no confidence in this product.

For tobacco use only

The really mind-blowing thing is that this has been in the term since October 2025. That doesn't excuse their abhorrent advertising practices, which are commonplace among big AI companies, but it's incredible nobody saw it for so long.

Lots of companies use this in legal documents just to absolve themselves of responsibility.

The article seems to be focussing on the "making bad decisions" aspect - i.e. "don't use it for anything important".

I wonder if this is also an attempt to limit IP liability in case someone claims that copilot reproduced copyrighted/patented material?

Obviously entertainment is also full of copyrighted material but the payouts aren't usually quite as big as patent claims.

It's just a game, but we're gonna put it in EVERYTHING! Play our game peasants!