[Ace17]

This is true in theory. Unfortunately, it doesn't prevent people from patenting algorithms in practice.

There's an interesting loophole: just also include in your invention the machine (memory + CPU) running the algorithm. Then you're patenting a device, instead of simply math.

Stallman described this method more than 10 years ago. I was hoping this wasn't true, until the (french) company I work in successfully used it to patent an underlying algorithm from the software I had written.

Maybe such a patent would be debunked in an european court? I'd be glad if this were the case.