Moderator and Panellists
- Steven Chia, Presenter and Senior Editor, CNA
- JJ Geewax, Director of Applied AI, Google DeepMind
- Mark Chen, Director of Trust & Safety Group (Anti-Scam Products Team), GovTech Singapore
- Justin Wong, Senior Assistant Commissioner of Police (SAC) Cyber Command, Singapore Police Force (SPF)
- Arifah Sharifuddin, Institute Director, Tech For Good Institute
- Raju Nair, Managing Director of Legal & Compliance (PASO), DBS
As scams grow increasingly sophisticated with artificial intelligence (AI), users face growing challenges in responding to AI-enabled scams. This concern is prevalent in Southeast Asia, where 60% of the population encounters scams, and approximately a third of these involve AI-generated impersonation. Such vulnerability extends beyond initial detection to how users respond once targeted. An upcoming study by TFGI found that only a minority of respondents know how to disengage safely, with many inadvertently worsening their exposure by maintaining communication with perpetrators.
Against this backdrop, there is a need to bring anti-scam efforts to the top of national agendas so as to ensure that victims have clear support mechanisms throughout the scam lifecycle. The panel highlighted that governments and financial institutions cannot safeguard the digital ecosystem alone. Instead, scam resilience requires shared responsibility, where users are actively equipped and empowered to protect themselves.
Ultimately, the panellists remained cognisant that technology alone cannot solve a fundamentally behavioural challenge. The human angle must be addressed by strengthening psychological resilience, adopting human-centric verification practices, and mobilising community networks.
Key takeaways
1. Society must build robust safeguards addressing both technological vulnerabilities and psychological manipulation
AI is fundamentally changing the nature of scams. In the past, spelling and grammatical errors served as primary markers of suspicious activity. With the advent of AI, however, convincing rhetoric and deepfakes impersonating public figures have rendered these traditional signals ineffective for scam detection. Simultaneously, scams have grown increasingly sophisticated, relying on an elaborate psychological scheme of attraction, persuasion, and execution. As such, it was unsurprising that 64% of respondents expressed growing concern and worry over the increasing difficulty of detecting scams.
Given the dual role of technology and psychology in driving scams, TFGI’s upcoming study offers critical insights. Across the region, 42% of respondents exhibited underconfidence, reporting a low ability to identify scams despite achieving a 71% accuracy rate in actual detection tests. As such, initiatives that help underconfident users feel more equipped against scam threats are essential, like GovTech’s scam vaccine, which inoculates the population and heightens readiness against emerging scam threats.
Nonetheless, the panel cautioned against over-reliance on AI countermeasures, lest the response turn into an arms race of competing models. For instance, results from AI verification of deepfakes can be misleading if the videos are genuine but misrepresented for fraudulent activities, a human nuance where AI may still falter. Thus, safeguards should be built around the user experience to mitigate the likelihood of victims buying into the psychological manipulation tactics of scammers.
2. Upstream digital safeguards must be supported by downstream methods like introducing friction at key touchpoints and human-centric verification.
In a society where digital transactions and activities are seamless, there is a growing expectation of convenience. The hidden cost of this convenience is reduced deliberation: shorter decision windows leave users more prone to decision-making based on emotional cues instead of rational ones. This vulnerability is further compounded by high societal trust, which discourages people from questioning what is presented to them. Consequently, this environment is one of the reasons why Singapore reports disproportionately high scam figures and financial losses compared to other jurisdictions.
Given that 80% of all scams are authorised by the victims themselves, anti-scam efforts must adapt to counter this industrialised deception. The first layer of defence requires detection to move upstream before fraudulent messages even reach users. As highlighted by the panel, this involves implementing cross-sector collaborations such as the National Scam List to intercept scam threats swiftly.
However, such initiatives need to be reinforced by downstream user interventions. Introducing intentional cognitive friction at key touchpoints across the user journey is particularly salient when an offer appears too good to be true. At the same time, instead of relying excessively on technology, solutions should also incorporate human-centric verification methods. For instance, agreeing on a safe phrase offline can serve as a trust signal, or requesting to meet in person offers a physical check that digital tools cannot manipulate. Thus, empowering individuals to protect themselves ensures that digital safety is not completely offloaded to institutions.
3. Combating increasingly convincing scams requires system-level interventions that serve as a failsafe for vulnerable groups
There exists a tension between individual autonomy and digital safety. Panellists highlighted cases where victims were so deeply entrenched in a scam narrative that financial institutions remained the final line of defence before funds were transferred. In this context of increasingly convincing scams, stronger legislation like the Protection from Scams Act is required to grant authorities the mandate to step in. While contentious, this framework empowers the police to issue timely restriction orders on a victim’s banking facilities while authorities investigate.
Such intervention proved essential in the case of a vulnerable woman who believed Elon Musk urgently needed her savings because he was facing financial troubles. Notably, she persisted in this belief even after her family attempted to dissuade her. Over the six-month period, the maximum duration a restriction order can remain in force, the victim was also provided with psychological and counselling support. This example demonstrates why law enforcement must collaborate closely with banks to prevent irreversible losses. Such synergy is also realised through the Anti-Scam Command, which centralises resources and improves capabilities across the ecosystem. Ultimately, safeguarding the public requires comprehensive end-to-end interventions that span upstream to post-scam support.
