Palo Alto Networks Reveals Potential Threats From AI Agentic Technology

Illustration of AI (photo: Freepic)

JAKARTA - Welcoming National Technology Awakening Day (Haktknnas) which is celebrated every August, the Palo Alto Networks emphasizes the importance of awareness of the rapid development of AI Agentics in Indonesia.

This advanced technology is even said to have capabilities that go beyond traditional generative models with the ability to execute, plan, and carry out actions autonomously.

"We need to highlight that one of Indonesia's digital destinys is determined by how we respond to and adopt AI Agentics, not just technological improvements; this is a paradigm change," said Adi Rusli, Country Manager, Indonesia, Palo Alto Networks.

In various involvements of the Palo Alto Networks with organizations in the region, a number of risks need to be watched out for by business people, including:

Order Text Injection (Prompt Injection) & Destination Manipulation: Hackers can enter malicious commands that divert AI systems from their duties or force them to leak confidential data, even with general instructions.

Abuse of Unsafe Tools & Integration: Analytic System relies on various third-party applications or Application Programming Interface (API) to operate. If this connection is not equipped with adequate access control, hackers can take advantage of the AI system's authority to carry out illegal actions on the company's system.

Credential Exposure & Identity Incognito: Agents can accidentally access or leak sensitive credentials, such as API key, token, or login information, which can cause an escalation of access rights or unauthorized access to the system. In a multi-agent environment, identity disguise is an increasingly serious threat.

Long Distance Code Execution (RCE): When an AI agent is given access to a code interpreter or script tool, malicious instructions can be inserted to run any code.

"To unlock their potential and secure Indonesia's digital future, security must be the foundation of trust and control," concluded Adi.