The Rise of Agentic AI: Opportunities and Security Challenges

Agentic artificial intelligence (AI) represents a significant evolution in automation and decision-making capabilities. Unlike traditional AI, which reacts to predefined inputs, agentic AI systems autonomously set goals, plan multi-step processes, and execute complex tasks. This shift promises enhanced efficiency for organizations, with a projected market size reaching approximately $197 billion by 2034, as noted by Terminal 3. However, this rapid growth also introduces substantial security vulnerabilities that many organizations are unprepared to address.

AI Agents: Emerging Threat or Opportunity?Snyk launches Evo for securing AI-native applicationsForbesWhy AI needs an identity: Gary Liu explains the first step in data securityLenovo unveils agentic AI to power the next generation of AI-enabled workforcesSingapore unveils efforts to govern agentic AI, prepare for post-quantum eraThe Straits Times

Despite the potential benefits, a survey by PwC revealed that while 79% of senior executives report AI agent adoption, only 44% have established security policies tailored to counter agent-related threats. This gap highlights a critical oversight, as agentic AI systems can create new attack surfaces and security challenges that extend beyond traditional paradigms. The Coalition for Secure AI has identified fifteen distinct threat vectors, including identity spoofing and agent communication poisoning, which pose real risks to organizations.

AI Agents: Emerging Threat or Opportunity?Snyk launches Evo for securing AI-native applicationsSingapore unveils efforts to govern agentic AI, prepare for post-quantum eraThe Straits TimesForbesWhy AI needs an identity: Gary Liu explains the first step in data securityRubrik Agent Cloud speeds enterprise AI with built-in security and guardrails

In response to the challenges posed by agentic AI, Singapore is taking proactive measures to govern its use. Digital Development and Information Minister Josephine Teo announced new initiatives aimed at addressing the risks associated with agentic AI and quantum computing. The Cyber Security Agency of Singapore (CSA) has released updated guidelines for securing AI systems, focusing on the unique risks of agentic AI. These guidelines are open for public consultation until December 2025, emphasizing the need for collective action to govern these technologies effectively.

Singapore unveils efforts to govern agentic AI, prepare for post-quantum eraThe Straits TimesAI Agents: Emerging Threat or Opportunity?Why AI needs an identity: Gary Liu explains the first step in data securitySnyk launches Evo for securing AI-native applicationsForbes

To mitigate the risks associated with agentic AI, experts recommend adopting disciplined engineering habits. These include constraining memory to task-specific limits, ensuring easy and thorough deletion of data, and implementing temporary, task-specific permissions for device access. Additionally, maintaining a readable "agent trace" can enhance transparency, allowing users to track what data is collected and when it will be deleted. By adhering to these practices, organizations can preserve functionality while minimizing their data footprint.

Agentic AI Security: Hidden Data Trails Exposed

Despite the advancements, there are concerns regarding the adequacy of current regulations and the pace of technological evolution. Gary Liu, CEO of Terminal 3, emphasized that existing regulatory frameworks are struggling to keep up with the rapid development of agentic AI. He warned that once data becomes part of an agent's memory, it is effectively irretrievable, raising critical questions about data governance and control.

Why AI needs an identity: Gary Liu explains the first step in data security

As agentic AI continues to evolve, the balance between leveraging its capabilities and ensuring robust security measures will be crucial. Organizations must prioritize the establishment of comprehensive security frameworks and governance structures to address the unique challenges posed by these systems. The proactive steps taken by Singapore serve as a model for other nations, highlighting the importance of international cooperation and timely action in navigating the complexities of agentic AI.

AI Agents: Emerging Threat or Opportunity?