ALMOST every telco I meet in Southeast Asia wants AI inside its app this year. What stalls these projects is rarely whether the technology works. It is the question the operator’s own security team asks first: How do we know what this thing is doing with our customers?
The vendors who get stuck answer, “Trust us.” The ones who get through answer, “Test us”: Run it on real traffic, see every signal it reads, switch it off whenever you like.
In the Philippines, that difference matters more than almost anywhere. Filipinos are quicker than their neighbors to abandon a service after a serious breach, and the AI law now being drafted would leave most everyday AI outside its safety checks.
Filipinos want AI
The Philippines has more mobile connections than people, 137 million in late 2025, according to DataReportal, and almost all of its online life runs through the phone. Scammers have followed. GSMA research released in November 2025 found that 52 percent of Filipinos have been scammed at least once, seven points above the Asean average, and that social media has overtaken text messages as the leading channel for scams. As Julian Gorman, head of Asia-Pacific at GSMA, put it: “The Philippines’ digital economy depends on trust. Our latest evidence shows that trust is being eroded faster than it is being rebuilt.”
AI is arriving straight into that environment, and consumers are conflicted about it. A Vero and Kadence International study of more than 3,000 consumers across six Southeast Asian markets found AI to be the least trusted technology category surveyed, yet 71 percent expect to use it more over the next year. Filipinos were the least forgiving of all: 56 percent would drop a technology service immediately after a serious data breach.
When AI sits inside a telco’s app, a failure does not belong to the algorithm. It belongs to the operator.
Patterns, not names
The AI race is usually framed as a contest between models. For telcos, it is the wrong race. They will not out-build the model labs, and they do not need to. They own the place where millions of people will actually meet AI.
Winning there does not require more personal data. Picture a prepaid subscriber who tops up every Friday and plays mobile games late at night. A well-built system can offer her a gaming bundle at the right moment without ever knowing her name, her address or who she messages. It reads patterns, not names.
The National Privacy Commission has already made clear that the Data Privacy Act applies to AI systems, including their training and testing. Privacy has to be designed in from the start, not bolted on at the compliance review.
Make ‘test us’ the standard
An operator should be able to run any AI system against real traffic before customers see it, know exactly which signals it uses and switch it off without an engineering project. Anything less is a black box with the operator’s logo on it.
Regulation will not require most of this. The House’s draft AI bill reserves registration and safety checks for high-risk systems that decide on loans, jobs and health. A recommended gaming bundle would not come close. That is sensible lawmaking, but it leaves the standard for everyday AI to the companies deploying it.
So telcos should set it, together. A shared, published baseline covering who tests a system before launch, what data it can see, where that data is processed and how it is switched off would let customers and regulators compare providers. Operators would compete on proof rather than promises.
I would rather the industry write that baseline now than have each operator learn why it matters after a breach.
Manit Parikh is founder and chief executive of The Binary Holdings, a technology infrastructure company that embeds Web3, blockchain, and AI-driven engagement layers directly into major telecommunication companies, e-wallets, and super-apps in emerging markets.
