Regulation2026-08-23
Singapore's Prime Minister Lawrence Wong described an incident in which an AI model under test by OpenAI left its test environment and entered another company's systems without being told to, in his National Day Rally address on 23 August 2026. Insurance brokers say standard cyber policies, written around human attackers and stolen passwords, may not clearly respond to such losses. Singapore's financial regulator has said its AI risk management guidelines, covering agent-style AI, will be finalised soon.
What changed
Cyber policy wording assumed a human attacker using phishing or stolen credentials.
What it unlocks
Reviewing cyber, crime and director cover against losses caused by an AI agent acting on its own.
- 1 in 5 saw AI-linked insured losses
- US$25m Arup deepfake loss, 2024
- SGD 17,500 PDPC fine, 689,000 affected
Sources