OpenAI’s Adversary Threat Report should be a prelude to more robust data sharing in the future. When it comes to AI, independent researchers are starting to compile misuse databases, such as the AI Incident Database and the Political Deepfake Incident Database, allowing researchers to compare different types of misuse and track how misuse changes over time. . However, misuse is often difficult to detect from the outside. As AI tools become more powerful and more common, it is important for policymakers considering regulation to understand how they are used and abused. OpenAI's first report provided a high-level summary and curated examples, but expanding our data-sharing relationships with researchers to provide greater visibility into hostile content or behavior is an important next step.
Online users also have a role to play in the fight against influence peddling and misuse of AI. Ultimately, this content only has impact when people see it, believe in it, and even engage in sharing it. In one of the cases disclosed by OpenAI, online users pointed out a fake account using AI-generated text.
In our own research, we found that the Facebook user community actively calls out AI-generated image content created by spammers and scammers, helping those who are less tech-savvy to avoid being tricked. Healthy skepticism is becoming increasingly useful. Helping friends and family become more aware of the growing prevalence of generated content, including pausing to check if the content is real and whether people are who they claim to be, will help social media users resist fraudulent activity. can. Both propagandists and fraudsters.
OpenAI's blog post announcing the takedown report succinctly stated, “Threat actors operate across the Internet.” So do we. As we transition into a new era of AI-driven influence operations, developing more resilient digital ecosystems requires addressing shared challenges through transparency, data sharing, and collaborative boundaries.
Josh A. Goldstein, Researcher He works on the CyberAI project at Georgetown University's Center for Security and Emerging Technologies (CSET). Rene Diresta Research Manager at Stanford Internet Observatory Invisible Rulers: Those Who Turn Lies into Reality.