GovTech and Google DeepMind developing real-time detection of AI scams on devices

Singapore's Government Technology Agency (GovTech) and Google DeepMind are teaming up to develop real-time detection of AI scams on devices, according to Minister for Digital Development and Information Josephine Teo. The collaboration aims to protect against AI-driven scams while maintaining privacy, as devices can identify early signs of manipulation to prevent financial or credential loss for users.

Teo emphasized that the fight against scams has become increasingly challenging due to AI usage by scammers. Instead of avoiding AI, Singapore must strengthen its defenses against AI-enabled scams. Scams often follow a multi-stage process, involving social media, messaging platforms, phone calls, and financial transactions. Each stage presents an opportunity for intervention, but no single organization can monitor the entire chain.

Therefore, connecting defences is crucial to trigger actions based on signals detected across the ecosystem.

AI can also be used to detect threats at scale. For example, the Scam Analytics and Tactical Intervention System developed by GovTech and the Singapore Police Force scans and assesses over 600,000 websites daily, disrupting about 20,000 malicious websites each month. This system enables Singapore to respond to threats rapidly and efficiently, surpassing the capabilities of human effort alone.

The minister also highlighted the need to prepare for quantum-related risks, as sufficiently powerful quantum computers could eventually break some of the cryptography protecting sensitive transactions and communications. While the timing of this potential threat is uncertain, Singapore must be prepared. Large organizations may take several years to migrate to quantum-safe solutions, involving risk assessment, prioritization, and gradual system migration.

To support this transition, a new quantum-safe kickstarter grant will provide funding to organizations to move from planning to implementation. The grant, open to all organizations with priority given to critical information infrastructure systems, will run from November 1, 2023, to October 31, 2027, with each pilot lasting up to six months.