Agentic AI introduces new risks and rewards for cybersecurity

From prompts to autonomous action

Traditional artificial intelligence typically functions by responding to specific human prompts. Agentic AI represents a shift toward systems that can analyze information, make independent decisions, and take actions with minimal human intervention.

Organizations are increasingly adopting these agents to automate routine tasks and improve decision-making speeds. Because these tools can act on their own, the cybersecurity challenge is expanding beyond protecting infrastructure to governing the AI itself.

As adoption of these autonomous systems gathers pace, companies must consider how to monitor and secure the agents. Even if fully autonomous offensive AI is not widespread yet, the technology is evolving fast enough to require immediate safeguards.

Accelerating threat detection

On the defensive side, agentic AI is reshaping how security teams identify and respond to cyber threats. In modern security operations centers, analysts are often overwhelmed by a high volume of alerts and false positives that can take days to validate manually.

Agentic AI can independently query data and analyze logs to present a complete picture of potential threats. As reported by KLSE Screener, this capability can reduce the time required for investigations from days down to minutes.

Unlike basic automation, these agents can build context over time to prioritize critical alerts. This allows them to filter out information overload and present clear findings to human defenders for swift, informed decision-making.

New risks of autonomy

The same autonomy that makes agentic AI a powerful defender also introduces significant technical and operational risks. Because these agents require privileged access to sensitive data and systems to function, their digital credentials become high-value targets for attackers.

If these credentials are not properly stored and monitored, they could be exploited to gain unauthorized access. Furthermore, an agent operating outside of defined parameters might reach the wrong conclusion or act too quickly in a crisis.

For instance, an AI might instantly isolate malware but accidentally shut down a server supporting medical life-support or financial transactions. Without carefully modeled guardrails, such autonomous errors can lead to serious real-world consequences.

Governance and human oversight

To manage these risks, Malaysia is strengthening its approach with the proposed Artificial Intelligence Governance Bill and the Cyber Security Act 2024. These measures reflect a growing need for clear accountability, transparency, and incident reporting in AI adoption.

Experts suggest that keeping humans in the loop remains non-negotiable for high-impact decisions. While agents can handle background processing, human judgment is still essential for complex incidents and strategic oversight.

Security professionals may see their roles shift as AI handles routine investigations. This allows them to spend more time on proactive threat hunting and complex tasks where human intuition is still required for success.

A phased path to adoption

Organizations should focus on a phased deployment approach to build confidence while minimizing operational risk. Starting in non-critical environments allows teams to test and refine AI systems before they are deployed in wider, more sensitive areas.

Success in this transition depends on finding professionals who can objectively assess the capabilities and limitations of the technology. Companies must also maintain fundamentals like system hardening, patch management, and protecting crown-jewel data.

Ultimately, agentic AI should not replace accountability. Organizations that combine autonomous technologies with strong cybersecurity fundamentals and experienced talent will be best positioned to manage the risks of this new era.

Collaborative security frameworks

Establishing practical safeguards for agentic AI requires more than just individual company policies. It demands closer collaboration between governments, technology providers, and businesses to create frameworks that can evolve alongside the technology. This collective approach ensures that security standards remain relevant as AI capabilities continue to advance.

Trusted platforms for joint research and real-time intelligence sharing are also essential for accelerating responsible adoption. By sharing information, the industry can reduce the risks of fragmented approaches and create a unified front against emerging cyber threats. This cooperation helps ensure that security measures are consistent across different sectors.