BANGKOK – Meta has published its latest Adversarial Threat Report, sharing how the company is tackling large-scale online scams across the Asia-Pacific (APAC) region and responding to a rise in AI-driven criminal activity. The report highlights close cooperation with law enforcement agencies in several countries.
In a Zoom interview with the Bangkok Post from the United States, David Agranovich, Meta’s Director of Security Policy, explained that the biannual report tracks major global trends in scams, influence campaigns, and other abuses that are increasingly powered by artificial intelligence.
A major section of the report covers the shutdown of criminal networks, many based in Cambodia, that posed as government officials and police officers. These groups targeted victims across Apac, including in Australia, Indonesia, the Philippines, Thailand, and Vietnam.
Meta notes that cross-border scam networks in the region have grown sharply in both scale and sophistication. These operations now use complex tactics, such as moving victims onto encrypted messaging apps and using alternative payment options that are harder to trace. Scammers rely on stolen or fake identities that look local and credible, then run scripted playbooks and automation tools to reach more people.
AI Deepfakes and Personalized Scams
According to Agranovich, these scam tactics appear over and over in countries like Australia, Singapore, and India. Impersonation scams and investment fraud remain the most common problems reported from these markets.
Criminals now use AI to make scams more convincing across Apac. This includes deepfake videos that copy the faces and voices of public figures, a trend that shows up most in Australia and India. AI tools also help scammers personalize outreach in bulk, tailoring messages so victims are more likely to respond.
Agranovich shared that Meta is deploying advanced behavioral detection systems and alerts for suspicious conversations. The company has also added new AI safety measures, including Llama Firewall and the Rule of Two framework for AI agents, to better protect people on its platforms.
Meta has been working with police in Singapore to support arrests linked to organized scam networks, as part of broader cooperation with law enforcement across the region. To reduce impersonation, the company has invited nearly 500,000 public figures to join its facial recognition program, with strong uptake in Australia, Singapore, and India, where this problem is especially common.
At the same time, Meta has taken down large numbers of fake accounts and is coordinating with law enforcement, banks, and other partners. This includes work with initiatives such as the US Department of Justice’s Scam Center Strike Force.
Large-scale Facebook Account Takedowns
From January to October, Meta removed more than 6,400 Facebook accounts and pages that broke its rules on fraud, scams, and dangerous individuals or organizations, Agranovich said. This network was linked to Cambodia and pretended to be officials and police from the United States, Australia, Indonesia, the Philippines, Thailand, and Vietnam.
He explained that the investigation sought to disrupt criminal networks, with special attention to so-called “recovery” or “re-scam” schemes. These scams target people who have already lost money, then pretend to be authorities who can help recover funds, for a fee.
According to the report, scammers in Indonesia, the Philippines, Thailand, and Vietnam used official-looking logos and visuals, such as insignia, badges, and images of uniformed officers. Some created fake organizations with names like “Cyber Crimes Investigations Service” or “Counter Fraud Alliance” to sound credible.
These groups posted publicly or joined Facebook groups and forums where victims were asking for help. They then moved conversations to chat apps and charged various fees under the claim of investigations or refunds. Victims never received any real assistance.
Four Pillars of Meta’s Anti-scam Strategy
Agranovich outlined four main pillars that guide Meta’s work against cyber threats.
1. Strengthening platform defenses
The first pillar focuses on building stronger systems, so it is harder for malicious actors to operate. This includes wider use of facial recognition for public figures, which helps prevent criminals from using their images in scam ads. According to Agranovich, expanding this technology has doubled Meta’s ability to find and remove such advertisements.
2. Empowering users with tools and education
The second pillar centers on giving people better tools, controls, and information. On Messenger, Meta is testing advanced scam detection features that alert users when a new contact sends a suspicious message. Users can send the message to AI for review and, if it appears to be a scam, receive clear guidance on how to block or report the account.
3. Disrupting and deterring scam networks
The third pillar focuses on breaking up entire malicious networks, both on Meta’s platforms and beyond. The company uses deep investigations and automated systems to spot and remove whole clusters of scam activity.
In 2025 alone, Meta removed more than 134 million scam advertisements. In the first half of that year, the company also disrupted nearly 12 million accounts linked to organized scam centers. According to Meta, user reports of scam ads dropped by more than 50 percent over the past 15 months.
4. Working with the broader defender community
The fourth pillar is about collaboration with partners, including industry groups, governments, and law enforcement agencies. Meta shares threat intelligence and publishes findings from its investigations.
One example is its work with the Singapore police, which helped break up an illegal online gambling network. This cooperation led to arrests and the recovery of more than US$175,000 (about 5.5 million baht) for victims.
New protections against AI-powered scams
Agranovich also described fresh defenses against AI-based threats. Criminals are now using AI tools to generate deepfakes for fake job offers, romance scams, and other frauds that rely on trust.
Meta is responding by using AI on the defensive side as well. Its systems are designed to detect patterns of abuse, flag questionable content and accounts, and give users timely alerts so they can protect themselves before money or data is lost.
