(photo by Matt Volpe)
Researchers at the University of Toronto have amassed hundreds of signatures on an open letter calling for improved AI safety protocols and oversight, warning that recent high-profile security incidents could foreshadow far more damaging failures without intervention.
The Open Letter on AI and Cybersecurity from the Scientific Community has been signed by nearly 500 scientists from 39 countries – including U of T University Professor emeritus and Nobel Prize winner Geoffrey Hinton and Professor Yoshua Bengio of the Université de Montréal, who are both widely considered “godfathers of AI.”
It calls for more research into AI safety and stronger measures to contain AI systems during testing, citing recent incidents involving AI agents acting unpredictably and “irresponsible” lapses in standard security practices. Recommendations include creating an international body to set safety standards and oversee the mandatory reporting of security incidents.
“Without immediate changes in AI development and deployment practices, such incidents will continue to occur, and future incidents will be far more damaging to our exposed critical infrastructure,” the letter reads.
The letter has made international headlines, garnering coverage in Reuters, Al Jazeera, Global News and the Canadian Press, as well as outlets based in Australia, Italy, Germany, France and Spain.
The initial draft was penned by U of T’s David Lie, director of the Schwartz Reisman Institute for Technology and Society, alongside Nicolas Papernot, associate professor of electrical and computer engineering in the Faculty of Applied Science & Engineering, and Thomas Ristenpart, professor of computer science in the Faculty of Arts & Science.
They shared it with colleagues to sign and pass on. “It just snowballed,” Lie told the Globe and Mail.
Ristenpart told the Globe that the letter’s widespread traction reflects the “joint effort” needed across academia and industry to conduct AI testing safely.
Almost two dozen U of T scientists are among the signatories, including Ron Deibert of the Citizen Lab at the Munk School of Global Affairs & Public Policy, as well as computer science researchers Alán Aspuru-Guzik, Sheila McIlraith, Zhijing Jin and Roger Grosse, who are each CIFAR AI Chairs at the Vector Institute for Artificial Intelligence.