DARPA Launches GARD Program to Defend AI Against Adversarial Attacks

Thumbnail Image

The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.

DARPA has selected Intel and Georgia Tech to lead the GARD program, a four-year, multimillion-dollar initiative to develop defenses against adversarial attacks on AI and machine learning systems. The effort aims to prevent future harms, such as misclassification in autonomous vehicles, by improving AI robustness against deception and data poisoning.[AI generated]

Why's our monitor labelling this an incident or hazard?

The event involves AI systems (machine learning models) and addresses the potential for harm through deception attacks that could lead to incidents like autonomous vehicle accidents. However, the article primarily reports on a research program aimed at preventing such harms rather than describing an actual harm event or realized incident. Therefore, it represents a credible effort to reduce plausible future harm from AI deception attacks, fitting the definition of an AI Hazard. It is not an AI Incident because no harm has yet occurred, nor is it Complementary Information since it is not an update or response to a past incident but a new initiative. It is not unrelated because it clearly involves AI systems and their security.[AI generated]
AI principles
Robustness & digital securitySafety

Industries
Government, security, and defenceMobility and autonomous vehicles

Affected stakeholders
ConsumersGeneral public

Business function:
Research and development

AI system task:
Recognition/object detectionEvent/anomaly detection


Articles about this incident or hazard