Agency leaders see vast potential in artificial intelligence, from applications of machine learning to newer solutions based on generative AI and now agentic AI. The problem is that these solutions also come with risks, especially in the areas of security, data privacy and AI bias. If agencies do not understand and address these risks up front, they leave themselves vulnerable to serious challenges in the future.
While the same dynamic has played out with every new technology, said Anhad Singh, Founder and CEO of Styrk AI, the increasingly rapid adoption of AI makes AI risks even greater.
“The focus now is on innovation and deregulation, and that’s great, because it enables people to just try it out,” said Singh. “But as always happens, the adversaries will catch up. So now the question is, do you want to leave yourself vulnerable, or do you want to be more proactive and make sure you cover the fundamentals as you’re adopting this?”
The risks are multifaceted, said Dr. Donnie Wendt, a researcher and author of the book, AI Strategy and Security: A Roadmap for Secure, Responsible, and Resilient AI Adoption. “When we start looking at AI, it’s really subject to various types of attacks throughout the entire lifecycle,” he said, “from the start, when we’re training, all the way through operationalization.”
In this video interview, Singh and Wendt discuss key issues that agencies must address to ensure safe and responsible AI. Topics include:
- Addressing concerns around safe and responsible AI from an initiative’s outset
- Mitigating agentic AI’s security risks
- Understanding the vulnerabilities of LLMs
