This is much easier said than done. DARPA moved on because they work to push the edge of research into proof of concepts. A lot of good work has been done to make 'robotic cars' a reality (sqrt-slam, better velodynes, deformable parts model for pedestrian detection, etc), but definitely doesn't mean i would trust a small team of hackers with my safety on the road.

Even the Google team, as much progress as they've made, is like 11-12 people working full time for the last few years to build something worth demonstrating. However, despite their resources and expertise, still have a lot to overcome. In talking with Jiajun Zhu (manager of the Google car team), it became readily apparent that they still encounter situations where they are not fully confident in their system (such as on streets with too much brush/foliage). Given the rocky start (in terms of security, features, and stability) that most start-ups have, i would never trust their finished product.