The Transportation Department is funding research to build artificial intelligence-powered drones and trying to get a better understanding of how autonomous vehicles could be manipulated to harm the public.
The agency’s Volpe National Transportation Systems Center on Thursday began soliciting teams to develop open-source artificial intelligence tools that allow drones to fly without a human operator. The team would ultimately build a drone capable of locating targets and dodging obstacles without outside intervention.
Selected vendors would also need to provide secure facilities where government engineers can store and test autonomous drones and cars. The 12-month contract is only open to companies and universities located within 25 miles of Arlington, Va., according to the solicitation.
While the vendors would primarily explore ways AI could augment drones, officials said they’re also curious how the tech might be manipulated to threaten the public.
“The dynamic and rapid advancement in, and ad hoc application of, machine learning in vehicle platforms … requires [the government] to quickly assess usage that individuals may attempt that could present unsafe conditions on our roadways, waterways and airspace,” officials wrote in the solicitation.
Bad actors can manipulate machine learning tools in many ways beyond exploiting basic cybersecurity flaws, and open source software is particularly prone to certain kinds of attacks. The agency said it’s currently researching how susceptible certain autonomous vehicles are to various types of threats and whether they can be co-opted to cause harm, but those efforts aren’t listed in the scope of the solicitation.
Responses are due Jan. 17.