Oak Ridge Lab Helps Set Community Standards for AI Security
Generative AI is increasingly becoming the backbone of many industries, powering services from next-generation chatbots like ChatGPT, Bard, and Llama, to critical sectors such as healthcare and finance. As these applications become deeply integrated into sectors dealing with sensitive data, there’s a growing call for accountability. Industry leaders, government officials, and academics are voicing concerns over the security of these AI models, emphasizing the need to safeguard them against potential threats like data tampering or unauthorized data access.
While the wheels of government regulation turn slowly, there’s a proactive movement within the cybersecurity community to establish defensive measures against these emerging threats. A notable contribution in this arena comes from Sean Oesch, a prominent figure in the field of emerging cyber technologies at the U.S. Department of Energy’s Oak Ridge National Laboratory (ORNL).
Oesch has played a pivotal role in the creation of the OWASP AI Security and Privacy Guide, an initiative aimed at setting global standards for AI security and helping to inform future legislation. “As AI models grow in capability, it becomes increasingly important to discuss and address the security and safety implications associated with their use,” Oesch stated. His affiliation with ORNL’s Center for AI Security Research underscores a deep commitment to advancing the safe and responsible application of AI technologies. His work focuses on assessing AI-driven cybersecurity solutions and engaging in critical conversations about AI’s security landscape.
The Open Web Application Security Project (OWASP) is renowned for its dedication to enhancing the security of software applications. It’s widely recognized for its influential top 10 lists that highlight common security vulnerabilities. The OWASP AI Security and Privacy Guide goes a step further, providing a comprehensive overview of potential AI system threats, detailing their impacts, and offering guidance on how they can be mitigated by information security professionals and data scientists. “This guide is designed to illustrate the attack vectors on AI systems, their consequences, and how these risks can be minimized,” explained Rob van der Veer, the founder of the OWASP AI Exchange. This initiative is a testament to the collective effort to bolster AI security and privacy in an era where AI’s influence continues to expand.
This collaborative spirit is crucial as AI technologies become more embedded in various industries. By sharing best practices and resources, the cybersecurity community is building a foundation upon which businesses can confidently integrate AI into their operations. The guidance provided by experts like Oesch and organizations such as OWASP plays an instrumental role in shaping a safer digital future.
UT-Battelle manages the Oak Ridge National Laboratory for the Department of Energy’s Office of Science. As the largest supporter of physical sciences research in the United States, the Office is addressing some of the most challenging issues we face today. For further information, visit the Department of Energy’s Office of Science website at energy.gov/science.
This press release aims to shed light on ORNL’s contributions to AI security standards, emphasizing the importance of collective action and the sharing of knowledge in addressing the evolving challenges within the field of artificial intelligence.
Note: The insights shared in this article are attributed to the individuals and organizations mentioned, with no institutional positions being endorsed. The perspectives provided are solely those of the corresponding authors.
