It's very harmless. Obviously there's a good number more technical details in here, but the gist of it is we take the parts of the HIV virus that makes it so good at proliferating, take out the genetic material that makes it harmful to our cells, and put in our own genetic information for the HIV-derived virus to inject into our target cells (known as a viral vector).
Contrary to the mental connotation, it's a very harmless procedure that's often used even in BSL-2 labs. It's not to say that there aren't risks involved, but chances are you're not going to end up with AIDs.
HIV-derived lentiviruses have been in use for a long time now and are a proven research tool. Coincidentally I just grew and harvested a new batch of lentiviruses this past Tuesday to treat cancer cells with shRNA.
In general, a virus consists of some genetic material and some method of getting that genetic material into a cell. You can swap out the virus's own genetic material for the genetic material of your choosing, and the delivery mechanism will still work. If you use HIV's "delivery system" to deliver genetic material that encodes a cancer-fighting protein, then you get cancer-fighting T-cells (a type of immune cell), since HIV targets T-cells.
This is a massive oversimplification, of course. But anyway, the harmlessness of the viral vector depends on how efficiently you can separate the virus's delivery mechanism from its own infections genetic material.
Nothing in life is without risk, and using HIV is no exception. However, in comparison to the risk of dying from leukemia, I think it is fair trade-off.
I wasn't aware such a thing exists. Anyone care to elaborate? Is it harmless or probably harmless?