Digital Safety Research Institute Helps Define AI Safety Terminology

As artificial intelligence (AI) sees wider deployment in the real world, the need for standardized definitions of “AI incidents” and “AI hazards” is increasingly important for risk management purposes.

Clearly defining these terms fosters international interoperability, while also allowing jurisdictions flexibility in determining the scope of AI incidents and hazards affecting their communities. To this end, the Organisation for Economic Cooperation and Development (OECD) published a recommendation on this topic titled “Defining AI incidents and related terms.” Two researchers from the Digital Safety Research Institute (DSRI) served as experts in its drafting.

Kevin Paeth, DSRI’s technical lead, and Sean McGregor, Ph.D., a DSRI director, are members of the OECD working group that informed the recommendation. They contributed their insights derived from developing the AI Incident Database.

“A shared definition is the start of sharing the development of safer AI systems,” McGregor said. “We are excited to move forward with a consensus definition for that which we aim to prevent – AI incidents.”

Learn More About the Framework