On 11 November 2026, cybersecurity firm Cequence Security and professional services giant KPMG unveiled a joint governance framework in Singapore, targeting one of the fastest-growing blind spots in enterprise technology: autonomous, or "agentic," artificial intelligence.

The new report, Agentic AI Runtime Security, Governance, and Control, highlights a critical shift in how artificial intelligence operates in the wild. Large language models have transitioned from passive text generators to active agents capable of executing multistep tasks. According to Cequence’s runtime data, these AI agents are already actively traversing enterprise networks—logging into accounts, booking services, and conducting financial transactions on behalf of users.

Crucially, these agents are frequently executing tasks without identifying themselves as non-human entities. For businesses in Aotearoa New Zealand and the wider Asia-Pacific region, this unverified activity presents a severe authentication and security crisis.

Traditional bot-defence systems were built to block malicious scrapers or automated credential-stuffing attacks. However, distinguishing between a legitimate human customer, a delegated AI assistant acting on that customer's behalf, and a malicious autonomous agent is proving difficult for standard application programming interface (API) gateways. This failure mode exposes organisations to complex financial liabilities, fraud, and data privacy breaches. If an AI agent hallucinates and makes a bulk purchasing error, or exceeds its delegated authority, the legal attribution and liability remain murky.

To harness the productivity benefits of agentic AI safely, the Cequence and KPMG collaboration advocates for a stringent operational baseline. The proposed framework focuses on how organisations can govern and secure agentic AI as models move from merely generating content to autonomously taking consequential actions. Establishing robust API governance is positioned as critical for ensuring that unverified automated activity operates within strictly enforced boundaries and does not exceed its delegated authority.

While the automation of complex workflows promises substantial efficiency gains for the digital economy, the unmonitored proliferation of agentic AI poses immediate systemic risks. As businesses across the Asia-Pacific race to integrate autonomous tools into their customer service and supply chain operations, establishing clear rules of engagement is no longer a theoretical exercise. Ensuring that every artificial agent operates transparently, securely, and within its designated authority has become an urgent prerequisite for maintaining digital trust.

Sources


AI disclosure: Automated systems assisted with research and drafting. Gemini independently peer-reviewed this article before publication. Sources are linked for verification.