Emerging AI Agents: A Mix of Security Concerns and Future Potential

AI agents have become a central focus in recent product launches, though cybersecurity experts are divided on their implications.

According to SailPoint data, 96% of tech professionals consider AI agents an increasing security threat. Nonetheless, nearly all respondents intend to expand their use of agentic AI in the upcoming year.

The main concern raised by these professionals pertains to the agents’ access to protected data, followed closely by the risk of unintended actions. Additionally, there are concerns about unauthorized sharing of sensitive information.

Data and Privacy

In generative AI platforms, similar issues have surfaced where models often make inaccurate or incorrect judgments. The ongoing black box challenge means that analysts frequently cannot ascertain why the AI made a mistake.

Privacy concerns are also significant for AI systems needing large datasets. Although well-known gen AI platforms like ChatGPT are designed to safeguard sensitive information, AI agents may still need access to personal details, such as financial data.

A concerning result from the SailPoint study indicated that approximately 25% of respondents reported their AI agents had disclosed access credentials.

Further findings showed that about 80% of participants observed their companies’ AI agents engaging in unintended activities, including unauthorized system access, dissemination of protected data, and retrieval of inappropriate content.

The Dawn of Agentic Commerce

In light of these concerns, the era of agentic commerce is rapidly evolving. Visa and Mastercard have introduced platforms that aim to transform AI agents into personal shoppers, allowing them to search for items and make purchases with minimal user intervention.

PayPal has also entered this space by teaming up with Perplexity to integrate payment functionalities directly within the AI platform’s chat interface.

The potential of AI agents is undeniable; many more initiatives are expected across various sectors, including cybersecurity. Yet, it is imperative that organizations maintain a strong focus on privacy and security in these endeavors.

According to the SailPoint study, 92% of respondents believe governing AI agents is crucial for enterprise security.