AI agents have featured prominently in recent product launches, but cybersecurity experts hold varied opinions on this technology.
A recent survey by SailPoint revealed that 96% of tech professionals perceive AI agents as an increasing threat to security. However, almost all respondents indicated they intend to expand their use of agentic AI in the coming year.
The primary concerns expressed by respondents included unauthorized access to sensitive data, unintentional actions, and the risk of sharing protected information without permission.
Data and Privacy
Similar issues arise with generative AI platforms where models often produce incorrect or misleading outcomes. Due to the persistent “black box” problem, analysts struggle to understand why AI makes errors.
Privacy concerns are also significant for AI models that rely on substantial data sets. While established gen AI platforms like ChatGPT are built with security measures, AI agents frequently need access to private information, including financial details.
A concerning finding from the SailPoint study was that nearly a quarter of respondents reported their AI agents had disclosed sensitive credentials.
Additionally, 80% of respondents noted instances where their company’s AI agents performed unintended actions, such as unauthorized system access and data dissemination.
The Age of Agentic Commerce
In spite of these worries, the era of agentic commerce is progressing. Visa and Mastercard have introduced platforms designed to transform AI agents into personal shoppers, enabling them to search for items and process payments with minimal user interaction.
PayPal swiftly followed by partnering with Perplexity to incorporate its payment services directly within the AI platform’s chat interface.
The potential of AI agents is undeniable, leading numerous initiatives to emerge across various industries, including cybersecurity. However, organizations must continually prioritize privacy and security in these endeavors.
This viewpoint was reinforced by the SailPoint study, which found that 92% of respondents considered governing AI agents crucial for enterprise security.