KUALA LUMPUR — Cybersecurity researchers, practitioners and industry leaders gathered at Parallel Pulse 2026 in Kuala Lumpur to examine emerging threats across artificial intelligence, threat intelligence, digital forensics, offensive security and enterprise defense.

The cybersecurity leadership initiative brought together speakers from across the region and beyond to share research, demonstrations and real-world experiences involving increasingly complex digital threats.

Organizers described the event’s philosophy through its name: “Parallel” represents the layered and synchronized defenses required in modern security environments, while “Pulse” reflects the continuous vigilance, intelligence and response needed to protect organizations.

AI agents become a new security battleground

One of the event’s key themes was the growing security implications of autonomous and agentic AI systems.

In his keynote, “Deception at Machine Speed: Building a Honeypot Strategy for the Agentic Era,” Laurent Oudot, Founder and CEO of OHIIHO in Singapore, presented research examining whether defenders can influence the behavior of offensive AI agents through the environments those agents interact with.

According to results presented during the session, controlled SSH evaluations found that 139 of 332 agent runs stopped completely after encountering an environmental security signal. The stops involved 81 of the 165 models tested.

A separate elicitation experiment found that 85% of 54 agents that completed the process disclosed the opening line of their system prompt, according to the presentation.

Oudot argued that defenders do not necessarily need direct control over an attacking model to influence its next decision. Instead, security teams can potentially use defender-controlled infrastructure and environmental signals to interrupt or influence agent behavior.

The research covered scenarios involving SSH and Linux environments, as well as other technologies including RDP/WinRM, PowerShell, web application attacks and basic scanning activity.

The findings were presented as specific to the models and configurations tested, highlighting the emerging possibility of using adversary interaction itself as a defensive mechanism.

Southeast Asian government data exposure investigated

Another session examined the infrastructure and operational mistakes behind a regional cyber threat.

In “One Hat Cyber Team: Infrastructure & SEA Government Data Exposure,” Sophia Pay, a researcher at Palo Alto Networks, presented an investigation into the One Hat Cyber Team and the exposure of Southeast Asian government-related data.

The investigation reportedly began with an open directory on a staging server. Researchers found customized web shells, attacker tools and exfiltrated Indonesian civil service records.

The exposed information reportedly extended beyond basic names and email addresses. By cross-referencing 17 structured datasets, investigators said it was possible to uncover information involving financial details, family relationships, employment histories and identity documents.

Operational security mistakes also provided investigators with additional clues. A screenshot of an operator’s desktop reportedly exposed information about compromised infrastructure, root access, personal conversations and links to other hacking communities in the region.

The case illustrated how exposed or vulnerable systems can become entry points for significant data compromise, even when attackers do not employ particularly sophisticated techniques.

Natural language emerges as an AI attack surface

The security implications of generative and agentic AI were explored further in “Natural Language, Unnatural Acts: Weaponizing the Agentic AI Attack Surface.”

Presented by Omkar Joshi, Security Architecture Manager at Coupa Software, and Pallavi Deshmukh, a cybersecurity leader and security researcher, the session demonstrated how seemingly ordinary interactions with AI systems could be manipulated.

The researchers showed scenarios in which an attacker could attempt to poison data through a support conversation before using incremental prompt changes to influence an AI agent and potentially extract sensitive information.

The demonstrations highlighted a growing concern for organizations deploying AI agents with access to internal systems, APIs and confidential information: natural language itself can become part of the attack surface.

Rather than replacing traditional vulnerabilities, AI agents can create additional pathways through which existing weaknesses may be reached or chained together.

Broad cybersecurity agenda

Parallel Pulse 2026 featured a wide-ranging speaker lineup covering several areas of modern cybersecurity.

Sessions included “Someone Else Turned On Your BitLocker” by Windows security researcher Sangsoo Jeong of 78ResearchLab in South Korea; “Double Agent: Can You Trust Your Security Agent?” by Savana Labs founder Ryan Teoh; and “Now You See Me: Deploying Your First AI Observability System” by Gregory Tan and Lancer Chua of U9up.

Other sessions included “The GReAT Ninja’s Guide: From IOC to Intelligence with OSINT & AI,” presented by security researcher Fareed Fauzi of Kaspersky, as well as talks focused on offensive security, threat intelligence and emerging cyber risks.

The event was supported by cybersecurity organizations and sponsors including Siaga Informatics, Kancil Research Labs, StealthMole and Netbytesec.

Security moves into the agentic era

The discussions at Parallel Pulse 2026 reflected a broader shift in cybersecurity: organizations are increasingly having to defend not only networks, endpoints and applications, but also AI systems capable of interpreting instructions, accessing data and taking actions autonomously.

From manipulating AI agents through natural language to using controlled environments to interrupt malicious automation, the research presented at the event explored both the risks and defensive possibilities created by the rapid adoption of agentic AI.

The sessions also reinforced a familiar lesson in cybersecurity: sophisticated technology is only part of the equation. Exposed infrastructure, poor operational security, vulnerable systems and human behavior can all contribute to major security incidents.

As organizations accelerate AI adoption, the challenge for security teams will increasingly be to understand how these new systems behave under attack—and how defenders can turn the same intelligence, automation and interaction capabilities into tools for protection.

Organiser and Partners

• Organiser: NanoSec.Asia

• Strategic Partner: SIAGA Informatics

• Gold Sponsor: Kancil Research Labs

• CTI Sponsor: Cyberwise

• Bronze Sponsor: NetByteSec Sdn Bhd

• Supporting organisations/partners: StealthMole, ConvergentDS, Ask Pentest and SINCON

Credit Image —  Nano Sec Asia