Introduction: Why Governance-First AI is Rewriting the Financial Crime Playbook

This article is the second instalment in our series, Governance-First AI Strategy: The Future of Financial Crime Detection. The series examines how financial institutions can move beyond box-ticking compliance and embrace AI systems that are transparent, trustworthy, and genuinely effective against crime.

If you missed Part 1 — The AI Governance Crisis: How Compliance-First Thinking Undermines Both Innovation and Compliance — we recommend it as a pre-read. There, we explored how today’s compliance-heavy frameworks have created a paradox: soaring costs, mounting false positives, and declining effectiveness in tackling sophisticated financial crime.

In this second part, we shift from diagnosing the crisis to highlighting solutions. We look at how governance-first AI is being operationalised through initiatives like Singapore’s AI Verify program, which is setting global benchmarks for validation, accountability, and continuous trust in financial crime detection.

The Governance Gap: Moving Beyond Checkbox Compliance

Traditionally, many financial institutions have seen governance as a final-layer exercise: a set of boxes to tick just before launching a new AML system or onboarding a new AI solution. But today’s complex, AI-driven systems have outpaced this outdated approach. Here’s why this gap is so dangerous:

The Risks of Outdated Governance

• Operational Failure: Financial institutions are reporting false positive alert rates reaching 90% or higher. Analysts spend valuable time on non-issues, while genuine risks can slip through unseen, creating an operational black hole.
• Regulatory Exposure: Regulators are increasingly sceptical of black-box AI systems that cannot be explained or audited. This raises the risk of costly penalties, strict remediation orders, and reputational damage.
• Stalled Innovation: The fear of non-compliance can make organisations hesitant to adopt even the most promising AI innovations, worried they will face issues during audits.

Towards Living Governance

True governance means embedding transparency, validation, and accountability across the entire AI lifecycle. This is not a static report, but a dynamic, ongoing protocol that evolves as threats and opportunities do.

AI Verify: Singapore’s Blueprint for Independent AI Validation

Enter AI Verify: Singapore’s response to the governance challenge, and a model now being emulated worldwide. Developed by the IMDA and AI Verify Foundation, this pioneering program aims to transform governance and validation from afterthoughts into core design principles for any AI system, especially those managing financial crime risk.

Key Features of AI Verify

• Rigorous, Scenario-Based Testing: Every AI model is evaluated against 400+ real-world financial crime detection scenarios, ensuring that outputs perform accurately across the range of complexities institutions actually face.
• Multi-language and Cross-Border Application: With testing in both English and Mandarin, AI Verify anticipates the needs of global financial institutions with diverse customer bases and regulatory environments.
• Zero Tolerance for Hallucinations: The program enforces strict protocols to ensure every AI-generated output is grounded in verifiable, auditable facts. This sharply reduces the risk of hallucinations, a key regulatory concern.
• Continuous Compliance Assurance: Validation is not a single event. Ongoing monitoring, reporting, and built-in alerts ensure the AI adapts to new criminal typologies and evolving regulatory expectations.

Validation in Action: The Tookitaki Case Study

Tookitaki became the first RegTech company to achieve independent validation under Singapore’s AI Verify program, setting a new industry benchmark for governance-first AI solutions.

• Accuracy Across Complexity: Our AI systems were validated against an extensive suite of real-world AML scenarios, consistently delivering precise, actionable outcomes in both English and Mandarin.
• No Hallucinations: With guardrails in place, every AI-generated narrative was rigorously checked for factual soundness and traceability. Investigators and regulators were able to audit the reasoning behind each alert, turning AI from a “black box” into a transparent partner.
• Compliance, Built-In: Stringent regulatory, privacy, and security requirements were checked throughout the process, ensuring our systems could not only pass today’s audits but also stay ahead of tomorrow’s standards.
• Strategic Trust: As recognised by media coverage in The Straits Times, Tookitaki’s independent validation became a source of trust for clients, regulators, and business partners, transforming governance into a strategic advantage.

Continuous Validation: Governance as Daily Operational Advantage

What sets AI Verify, and governance-first models more broadly, apart is the principle of continuous validation:

• Pre-deployment: Before launch, every model is stress-tested for robustness, fairness, and regulatory fit in a controlled, simulated real-world setting.
• Post-deployment: Continuous monitoring ensures that as new fraud threats and compliance rules arise, the AI adapts immediately, preventing operational surprises and keeping regulator confidence high.

This approach lets financial institutions move from a reactive, firefighting mentality to a proactive, resilient operating style.

The Strategic Payoff: Governance as a Differentiator

What is the true value of independent, embedded validation?

• Faster, Safer Innovation: Launches of new AI models become quicker and less risky, since validation is built in, not tacked on at the end.
• Operational Efficiency: With fewer false positives and more explainable decisions, investigative teams can focus energy where it matters most: rooting out real financial crime.
• Market Leadership: Governance-first adopters signal to clients, partners, and regulators that they take trust, transparency, and responsibility seriously, building long-term advantages in reputation and readiness.

Conclusion: Tomorrow’s AI, Built on Governance

As we highlighted in Part 1, compliance-first frameworks have proven costly and ineffective, leaving financial institutions trapped in a cycle of escalating spend and diminishing returns. AI Verify demonstrates what a governance-first approach looks like in practice: validation, accountability, and transparency built directly into the design of AI systems.

For Tookitaki, achieving independent validation under AI Verify was not simply a compliance milestone. It was evidence that governance-first AI can deliver measurable trust, precision, and operational advantage. By embedding continuous validation, institutions can move from reactive firefighting to proactive resilience, strengthening both regulatory confidence and market reputation.

Key Takeaways from Part 2:

1. Governance-first AI shifts the conversation from “being compliant” to “being trustworthy by design.”
2. Continuous validation ensures models evolve with emerging financial crime typologies and regulatory expectations.
3. Independent validation transforms governance from a cost centre into a strategic differentiator.

What’s Next in the Series

In Part 3 of our series, Governance-First AI Strategy: The Future of Financial Crime Detection, we will explore one of the most pressing risks in deploying AI for compliance: AI hallucinations. When models generate misleading or fabricated outputs, trust breaks down, both with regulators and within institutions.

We will examine why hallucinations are such a critical challenge in financial crime detection and how governance-first safeguards, including Tookitaki’s own controls, are designed to eliminate these risks and make every AI-driven decision auditable, transparent, and actionable.

Stay tuned.

In August 2025, Malaysian police stormed a five-storey office in Bangsar South, Kuala Lumpur, arresting more than 400 people linked to what is now called the country’s largest scam call centre operation.

The raid made headlines worldwide, not only for its scale but also because of its alleged link to Doo Group, a Singapore-based fintech that sponsors English football giant Manchester United. The case has cast a harsh spotlight on the industrial scale of financial crime in Southeast Asia and the reputational risks it poses for both financial institutions and global brands.

Background of the Scam

The dramatic raid took place on 26 August 2025, when Malaysian authorities swept into a commercial tower in Bangsar South, a thriving business district in Kuala Lumpur. Inside, they discovered a massive call centre allegedly set up to defraud victims across multiple countries.

Over 400 individuals were arrested. Videos of employees being escorted into police vans quickly went viral, symbolising the scale and industrial nature of the operation.

Initial reports linked the call centre to Doo Group, a global financial services provider with operations across Singapore, Hong Kong, London, Sydney, and Dubai. While the company has insisted that its operations remain unaffected and that it is cooperating fully with investigators, the reputational damage was already significant.

The Bangsar South raid is part of Malaysia’s wider anti-scam campaign. By mid-2025, authorities had arrested over 11,800 suspects in similar cases, with financial losses amounting to RM 1.5 billion (USD 355 million). The Bangsar South case, however, stands out because of its size, its international profile, and its link to a company with a global brand presence.

What the Case Revealed

The raid revealed troubling insights into how financial crime networks operate in the region:

1. Industrialised Fraud

A workforce of over 400 suggests this was not a small, fly-by-night scam but a structured enterprise. Staff were reportedly trained to follow scripts, handle objections, and target victims methodically, mirroring the efficiency of legitimate customer service operations.

2. Global Targeting

Reports indicate the call centre targeted victims not just in Malaysia but also overseas, raising questions about how funds were laundered across borders. The multilingual capabilities of employees further suggest international reach.

3. Reputation at Risk

The alleged connection to Doo Group highlights how reputable financial companies can be pulled into fraud narratives. Even if not directly complicit, the association underscores how thin the line can be between legitimate fintech operations and the shadow economy.

4. Oversight Gaps

The case also points to challenges regulators face in monitoring sprawling call centre operations and cross-border financial flows. By the time raids occur, thousands of victims may already have been defrauded.

Impact on Financial Institutions and Corporates

The Bangsar South raid is not just a law enforcement victory. It is a warning signal for the financial industry.

1. Reputational Fallout

When a Manchester United sponsor is linked to scams, it is not just the company that suffers. Brand trust in fintech, sports, and banking becomes collateral damage. This raises the stakes for due diligence in sponsorships and partnerships.

2. Investor and Customer Confidence

Digital finance thrives on trust. When fintechs are tied to scandals, investors hesitate and customers second-guess their safety. The Bangsar South case risks dampening enthusiasm for fintech adoption in Malaysia and the wider region.

3. Operational Risks for Banks

For financial institutions, call centre scams translate into suspicious transaction flows, mule account proliferation, and higher compliance costs. Traditional transaction monitoring often struggles to flag layered, cross-border flows connected to scams of this scale.

4. Regional Implications

Malaysia’s crackdown shows commendable resolve, but it also exposes the country as a hub for organised scam activity. This dual image, both a problem centre and an enforcement leader, will shape how regional regulators approach financial crime.

Lessons Learned from the Scam

1. Scale ≠ Legitimacy
   A large workforce and polished infrastructure do not guarantee a legitimate business. Regulators and partners must look beyond appearances.
2. Due Diligence is Non-Negotiable
   Global brands and institutions need deeper checks before partnerships. A sponsorship or corporate tie-up can quickly become a reputational liability.
3. Regulatory Vigilance Matters
   The Bangsar South raid shows what decisive enforcement looks like, but it also reveals how long such scams can operate before being stopped.
4. Cross-Border Cooperation is Critical
   Victims were likely spread across multiple jurisdictions. Without international collaboration, enforcement remains reactive.
5. Public Awareness is Essential
   Scam call centres thrive because victims are unaware. Public education campaigns must go hand-in-hand with enforcement.

The Role of Technology in Prevention

Conventional compliance methods, such as simple blacklist checks or static rules, are no match for scam call centres operating at an industrial scale. To counter them, financial institutions need adaptive, intelligence-driven defences.

This is where Tookitaki’s FinCense and the AFC Ecosystem come in:

• Typology-Driven Detection
  FinCense continuously updates detection logic based on real scam scenarios contributed by 200+ global financial crime experts in the AFC Ecosystem. This means emerging call centre scam patterns can be identified faster.
• Agentic AI
  At the heart of FinCense is an Agentic AI framework, a network of intelligent agents that not only detect suspicious activity but also explain every decision in plain language. This reduces investigation time and builds regulator confidence.
• Federated Learning
  Through federated learning, FinCense enables banks to share insights on scam flows and mule account behaviours without compromising sensitive data. It is collective intelligence at scale.
• Smart Case Disposition
  When alerts are triggered, FinCense’s Agentic AI generates natural-language summaries, helping investigators prioritise critical cases quickly and accurately.

Moving Forward: The Future of Scam Call Centres

The Bangsar South raid may have shut down one operation, but the fight against scam call centres is far from over. As enforcement improves, fraudsters will adopt AI-driven tools, deepfake impersonations, and more sophisticated laundering methods.

For financial institutions, the path forward is clear:

• Strengthen collaboration with regulators and peers to track cross-border scam flows.
• Invest in adaptive technology like FinCense to stay ahead of criminal innovation.
• Educate customers relentlessly about new fraud tactics.

The raid was a victory, but it was also a warning.

If one call centre with 400 employees can operate in plain sight, imagine how many others remain hidden. The only safe strategy for financial institutions is to stay one step ahead with collaboration, intelligence, and next-generation technology.

It began with a phone call. A senior citizen in Navi Mumbai answered a number that appeared to belong to the police. Within hours, she was trapped on a video call with men in uniforms, accused of laundering money for terrorists. Terrified, she wired ₹21 lakh into what she believed was a government-controlled account.

She was not alone. In August 2025, cases of “digital arrest” scams surged across India. An elderly couple in Madhya Pradesh drained nearly ₹50 lakh of their life savings after spending 13 days under constant video surveillance by fraudsters posing as investigators. In Rajkot, criminals used the pretext of a real anti-terror operation to extort money from a student.

These scams are not crude phishing attempts. They are meticulously staged psychological operations, exploiting people’s deepest fears of authority and social disgrace. Victims are not tricked into handing over passwords. They are coerced, minute by minute, into making transfers themselves. The results are devastating, both for individuals and the wider financial system.

Background of the Scam

The anatomy of a digital arrest scam follows a chillingly consistent script.

1. The Call of Fear
Fraudsters begin with a phone call, often masked to resemble an official number. The caller claims the victim’s details have surfaced in a serious crime: drug trafficking, terror financing, or money laundering. The consequences are presented as immediate arrest, frozen accounts, or ruined reputations.

2. Escalation to Video
To heighten credibility, the fraudster insists on switching to a video call. Victims are connected to people wearing uniforms, holding forged identity cards, or even sitting before backdrops resembling police stations and courtrooms.

3. Isolation and Control
Once on video, the victim is told they cannot disconnect. In some cases, they are monitored round the clock, ordered not to use their phone for any purpose other than the call. Contact with family or friends is prohibited, under the guise of “confidential investigations.”

4. The Transfer of Funds
The victim is then directed to transfer money into so-called “secure accounts” to prove their innocence or pay bail. These accounts are controlled by criminals and serve as the first layer in complex laundering networks. Victims, believing they are cooperating with the law, empty fixed deposits, break retirement savings, and transfer sums that can take a lifetime to earn.

The method blends social engineering with coercive control. It is not the theft of data, but the hijacking of human behaviour.

What the Case Revealed

The 2025 wave of digital arrest scams in India exposed three critical truths about modern fraud.

1. Video Calls Are No Longer a Guarantee of Authenticity
For years, people considered video more secure than phone calls or emails. If you could see someone’s face, the assumption was that they were genuine. These scams demolished that trust. Fraudsters showed that live video, like written messages, can be staged, manipulated, and weaponised.

2. Authority Bias is a Fraudster’s Greatest Weapon
Humans are hardwired to respect authority, especially law enforcement. By impersonating police or investigators, criminals bypass the victim’s critical reasoning. Fear of prison or social disgrace outweighs logical checks.

3. Coercion Multiplies the Damage
Unlike phishing or one-time deceptions, digital arrests involve prolonged psychological manipulation. Victims are kept online for days, bombarded with threats and false evidence. Under this pressure, even cautious individuals break down. The results are not minor losses, but catastrophic financial wipe-outs.

4. Organised Networks Are Behind the Scenes
The professionalism and scale suggest syndicates, not lone operators. From forged documents to layered mule accounts, the fraud points to criminal hubs capable of running scripted operations across borders.

Impact on Financial Institutions and Corporates

Though victims are individuals, the implications extend far into the financial and corporate world.

1. Reputational Risk
When victims lose life savings through accounts within the banking system, they often blame their bank as much as the fraudster. Even if technically blameless, institutions suffer a hit to public trust.

2. Pressure on Fraud Systems
Digital arrest scams exploit authorised transactions. Victims themselves make the transfers. Traditional detection tools that focus on unauthorised access or password breaches cannot easily flag these cases.

3. Global Movement of Funds
Money from scams rarely stays local. Transfers are routed across borders within hours, layered through mule accounts, e-wallets, and fintech platforms. This complicates recovery and exposes gaps in international coordination.

4. Corporate Vulnerability
The threat is not limited to retirees or individuals. In Singapore earlier this year, a finance director was tricked into wiring half a million dollars during a deepfake board call. Digital arrest tactics could just as easily target corporate employees handling high-value transactions.

5. Regulatory Expectations
As scams multiply, regulators are pressing institutions to demonstrate stronger customer protections, more resilient monitoring, and greater collaboration. Failure to act risks not only reputational damage but also regulatory penalties.

Lessons Learned from the Scam

For Individuals

• Treat unsolicited calls from law enforcement with suspicion. Real investigations do not begin on the phone.
• Verify independently by calling the published numbers of agencies.
• Watch for signs of manipulation, such as demands for secrecy or threats of immediate arrest.
• Educate vulnerable groups, particularly senior citizens, about how these scams operate.

For Corporates

• Train employees, especially those in finance roles, to recognise coercion tactics.
• Require secondary verification for urgent, high-value transfers, especially when directed to new accounts.
• Encourage a speak-up culture where staff can challenge suspicious instructions without fear of reprimand.

For Financial Institutions

• Monitor for mule account activity. Unexplained inflows followed by rapid withdrawals are a red flag.
• Run customer awareness campaigns, explaining how digital arrest scams work.
• Share intelligence with peers and regulators to prevent repeat incidents across institutions.

The Role of Technology in Prevention

Digital arrest scams prove that traditional safeguards are insufficient. Fraudsters are not stealing credentials but manipulating behaviour. Prevention requires smarter, adaptive systems.

1. Behavioural Monitoring
Transactions made under duress often differ from normal patterns. Advanced analytics can detect anomalies, such as sudden large transfers from accounts with low historical activity.

2. Typology-Driven Detection
Platforms like Tookitaki’s FinCense leverage the AFC Ecosystem to encode real-world scam scenarios into detection logic. As digital arrest typologies are identified, they can be integrated quickly to improve monitoring.

3. AI-Powered Simulations
Institutions can run simulations of coercion-based scams to test whether their processes would withstand them. These exercises reveal gaps in escalation and verification controls.

4. Federated Learning for Collective Defence
With federated learning, insights from one bank can be shared across many without exposing sensitive data. If one institution sees a pattern in digital arrest cases, others can benefit almost instantly.

5. Smarter Alert Management
Agentic AI can review and narrate the context of alerts, allowing investigators to understand whether unusual activity stems from duress. This speeds up response times and prevents irreversible losses.

Conclusion

The digital arrest scam is not just a fraud. It is a form of psychological captivity, where victims are imprisoned through fear on their own devices. In 2025, India saw a surge of such cases, stripping people of their savings and shaking trust in digital communications.

The message is clear: scams no longer rely on technical breaches. They rely on exploiting human trust. For individuals, the defence is awareness and verification. For corporates, it is embedding strong protocols and encouraging a culture of questioning. For financial institutions, the challenge is profound. They must detect authorised transfers made under coercion, collaborate across borders, and deploy AI-powered defences that learn as fast as the criminals do.

If 2024 was the year of deepfake deception, 2025 is becoming the year of coercion-based fraud. The industry’s response will determine whether scams like digital arrests remain isolated tragedies or become a systemic crisis. Protecting trust is no longer optional. It is the frontline of financial crime prevention.