I've read that ECoG's are being tested in prosthesis, which is very cool. I would think that a less invasive alternative (BCI is one, correct?) would become available before "brain implants" would be offered to the general public. The challenge as I understand it is regarding the sensitivity of the sensors, as well as the signal-to-noise ratio. Is there alot of development in this area, or is it limited to NASA?
Well in theory any Brain-Computer Interface is a BCI, whether it is based on EEG, ECoG, fNIR, or another technology, but I understand what you are asking and no, EEGs and fNIR are not considered invasive. At worst you are spending a lot of time for preparation (30 minutes or more when many electrodes are involved), getting conductive gel in the user's hair, or placing them inside a large machine for the duration of use.
Sensitivity of sensors is important, but removing noise and artefacts from ambience electrical devices and simple muscle movements (eyeblinks, twitch response, talking, etc.) are the greater issue. In addition you have to ensure that sensors are refitted with as much precision as possible in between sessions (a different position may yield slightly different signals or least the signals you used last time may be closer or farther away now), not to mention physical changes in the brain including neuron migration as a result of neuroplasticity - the actual act of training yourself to use the software produces changes in the brain to which the software needs to continue to adapt.
If you are interested to learn more, you may want to check out BCI2000 and OpenViBE, both of which are Open Source and produced by extremely well respected academic institutions.