Hackers could use AI to automate attacks, crash cars and drones

Rapid advances in artificial intelligence are raising risks that malicious users will soon exploit the technology to mount automated hacking attacks, cause driverless car crashes or turn commercial drones into targeted weapons, a new report warns.

AI could also generate images to impersonate others online and sway public opinion, report says

A man takes part in a hacking contest during the Def Con hacker convention in Las Vegas, Nevada, U.S. on July 29, 2017. A new report says AI could soon be used to mount automated hacking attacks. (Steve Marcus/Reuters)

Rapid advances in artificial intelligence are raising risks that malicious users will soon exploit the technology to mount automated hacking attacks, cause driverless car crashes or turn commercial drones into targeted weapons, a new report warns.

The study, published on Wednesday by 25 technical and public policy researchers from Cambridge, Oxford and Yale universities along with privacy and military experts, sounded the alarm for the potential misuse of AI by rogue states, criminals and lone-wolf attackers.

The researchers said the malicious use of AI poses imminent threats to digital, physical and political security by allowing for large-scale, finely targeted, highly efficient attacks. The study focuses on plausible developments within five years.

"We all agree there are a lot of positive applications of AI," Miles Brundage, a research fellow at Oxford's Future of Humanity Institute. "There was a gap in the literature around the issue of malicious use."

Artificial intelligence, or AI, involves using computers to perform tasks normally requiring human intelligence, such as taking decisions or recognizing text, speech or visual images.

It is considered a powerful force for unlocking all manner of technical possibilities but has become a focus of strident debate over whether the massive automation it enables could result in widespread unemployment and other social dislocations.

Hackers could use AI to cause driverless car crashes, the new report warns. (Syda Productions/Shutterstock)

The 98-page paper cautions that the cost of attacks may be lowered by the use of AI to complete tasks that would otherwise require human labour and expertise. New attacks may arise that would be impractical for humans alone to develop or which exploit the vulnerabilities of AI systems themselves.

It reviews a growing body of academic research about the security risks posed by AI and calls on governments and policy and technical experts to collaborate and defuse these dangers.

Impersonation threat

The researchers detail the power of AI to generate synthetic images, text and audio to impersonate others online, in order to sway public opinion, noting the threat that authoritarian regimes could deploy such technology.

The report makes a series of recommendations including regulating AI as a dual-use military/commercial technology.
 
It also asks questions about whether academics and others should rein in what they publish or disclose about new 
developments in AI until other experts in the field have a chance to study and react to potential dangers they might pose.

"We ultimately ended up with a lot more questions than answers," Brundage said.

The paper was born of a workshop in early 2017, and some of its predictions essentially came true while it was being written. The authors speculated AI could be used to create highly realistic fake audio and video of public officials for propaganda purposes.

Late last year, so-called "deepfake" pornographic videos began to surface online, with celebrity faces realistically melded to different bodies.

"It happened in the regime of pornography rather than propaganda," said Jack Clark, head of policy at OpenAI, the group founded by Tesla Inc CEO Elon Musk and Silicon Valley investor Sam Altman to focus on friendly AI that benefits humanity. "But nothing about deepfakes suggests it can't be applied to propaganda."

Comments

To encourage thoughtful and respectful conversations, first and last names will appear with each submission to CBC/Radio-Canada's online communities (except in children and youth-oriented communities). Pseudonyms will no longer be permitted.

By submitting a comment, you accept that CBC has the right to reproduce and publish that comment in whole or in part, in any manner CBC chooses. Please note that CBC does not endorse the opinions expressed in comments. Comments on this story are moderated according to our Submission Guidelines. Comments are welcome while open. We reserve the right to close comments at any time.