BSP flags cybersecurity risks from frontier AI models

BANKS and other financial institutions must ensure they have robust safeguards against potential threats from highly advanced artificial intelligence (AI) models as these could exploit vulnerabilities with a higher degree of sophistication, the Bangko Sentral ng Pilipinas (BSP) said.

In a memorandum signed by BSP Deputy Governor Lyn I. Javier, the BSP said frontier AI systems pose threats to the financial system as they can detect software vulnerabilities and exploitation routes, as well as perform multi-stage cyberattacks with minimal human intervention.

“While access to these systems remains restricted and controlled, their emergence signals a shift toward increasingly adaptive and scalable cyberthreats,” the central bank said.

“Accordingly, BSIs (BSP-supervised financial institutions) are reminded that similar AI-enabled capabilities may eventually be leveraged by malicious actors to target financial systems, third-party service providers, and critical infrastructure.”

Institutions should reinforce their cybersecurity and technology risk management frameworks to ensure these can address growing risks from cyberthreats driven by more advanced AI systems, the BSP said

“Sound practices for asset management, identity and access management, vulnerability management, security monitoring, incident response, and third-party risk management, among others, must be consistently implemented and maintained.”

Global banking regulators have raised concerns over advanced AI models like Anthropic’s Mythos, which cybersecurity experts have said could pose significant challenges to the banking industry and its legacy technology systems, Reuters reported.

The capabilities of Mythos to code at a high level have given it a potentially unprecedented ability to identify cybersecurity vulnerabilities, experts say.

Regulators globally have begun to focus more keenly on the impact of AI, from cyber and operational risks associated with frontier AI models to the challenges posed by the agentic systems capable of acting with limited human intervention.

Among the measures recommended by the BSP include having better visibility of their attack surfaces and integrating AI-enabled defensive tools for real-time threat detection and automated patching.

The central bank also urged BSIs to adopt micro-segmentation, implement zero trust architectures, and shift to multi-factor authentication using hardware security keys, smart cards, or hardware-backed certificate-based verification to counter AI-enabled social engineering.

Ms. Javier earlier said that social engineering schemes, including phishing scams, account takeover, and identity theft, were the top cyberthreat to the country’s digital payments system. These accounted for 76% of the total amount lost to financial fraud in the first half of 2025.

The BSP added that financial institutions must reassess and improve their Business Continuity Management framework and Business Continuity Plans.

“Finally, BSIs are recommended to formally develop their own AI Governance Framework, proportionate to the nature, extent, scale, complexity, and materiality of their AI systems, as well as the institution’s overall operational complexity and risk profile, following the principles set forth under BSP Memorandum No. M-2026-031 dated 24 June 2026,” the central bank said.

The BSP memorandum issued last month outlines guidelines for local financial institutions’ ethical AI adoption, which it said must be governed by the principles of sustainability, transparency, accountability, responsibility, and security. — Katherine K. Chan