Creating shared intelligence to stop telecom scams earlier

Stopping scams before they reach customers

Customer trust is vital for communications service providers. Fraudulent calls and messages impersonating banks, government agencies, logistical brands and digital services expose customers to financial loss and identity theft, while damaging confidence in mobile services.

Scam strategies change quickly as fraudsters alter content and sender identities, rotate numbers and move between SMS, MMS, voice and international routes. By the time a complaint is investigated, customers may already have been reached. Rule-based firewalls and manual analysis can miss new behavior.

The Catalyst project, TrustNet AI: Autonomous scam prevention for telecom networks, explores how operators can identify emerging threats and act earlier. It brings together adaptive AI-assisted detection, controlled response, and trusted intelligence sharing, while allowing each CSP to retain control of customer data, risk policies, and enforcement decisions.

Connecting detection across messages, calls, and networks

TrustNet AI uses specialized agents to analyze standardized SMS, MMS and voice events alongside subscriber context, sender information, network signals and reputation data. Behavioral and graph analysis connect related numbers, devices and recipients, revealing coordinated campaigns that may appear to be separate incidents.

The voice architecture supports caller-ID and spoofing checks, robocall and nuisance-call detection, Wangiri fraud detection, flash calls and voice impersonation. These capabilities complement existing firewalls and rules.

To support shared trust across operator boundaries, the Catalyst defines common scam-event formats, risk levels, reason codes, approved responses, and interfaces. A permissioned consortium ledger is proposed to hold attestations, number-rights status, messaging service provider registrations, and confirmed scam intelligence without placing raw customer records or traffic on the ledger. Signed trusted marks can point to ledger records and contribute to risk assessment alongside traffic analytics, roaming context, complaints, and verified intelligence.

AI agents investigate, score and explain; they do not enforce on the network alone. An orchestration step combines findings, evidence, and risk assessments before selecting an action from an approved catalog. Responses can include further investigation, alerts, customer warnings, redirects, or temporary and permanent blocks. Clear cases can be automated within approved policies, while uncertain or high-impact decisions require human review. Workflows manage validation, approval, enforcement, verification, expiration, and reversal including checks such as recent SIM swap or port before auto-block where configured.

Standards provide a common control model

TM Forum's Open Digital Architecture provides modular building blocks for event intake, AI-assisted detection, orchestration, case management and network enforcement, reducing the need for point-to-point integration.

TMF688 Event Management API standardizes scam events and outcomes, while TMF621 Trouble Ticket APIsupports cases and analyst updates. TMF633 Service Catalog API defines approved actions, and TMF701 Process Flow Management API coordinates validation, approval, enforcement and rollback.

TMF641 Service Ordering, TMF645 Service Qualification and TMF652 Resource Ordering are identified as production controls for checking and applying authorized network changes.

The CAMARA-aligned Scam Signal API gives approved banks and partner CSPs constrained risk indicators, like whether a call is active, exceeds a duration threshold, or is assessed as suspicious. This without exposing raw CDRs, message content or full communication history. TM Forum guidance on generative AI supports explainability, human oversight and controlled automation.

Targeting lower investigation costs and faster protection

TrustNet AI aims to give the industry a shared way to detect, assess, and respond to scams. Common event formats, reason codes, and workflows can reduce duplicated investigations, while a shared case record can help scam operations and contact center teams work from the same evidence. Each CSP continues to apply its own thresholds, approval policies, and enforcement decisions. Related alerts can be correlated into one campaign case so scam operations and contact centre teams work from the same evidence, with the scam desk as the system of record where integrated.

The project team targets a 25% reduction in scam investigation costs through the use of Agentic AI. It also aims to protect customer trust and enable new network API revenue. The wider intended outcomes include earlier intervention, lower operational effort, and more consistent responses across network boundaries.

Standard interfaces could help vendors integrate with multiple operators. Banks, payment providers and partner CSPs could receive consistent risk signals without raw network data, while regulators gain clearer decision and outcome records. This could reduce duplicated integration effort and improve consistency across the wider telecom ecosystem.

According to Sunil Gupta, E2E Network Architecture Director at Optus, “scams do not stop at network boundaries, but operator responses often do. TrustNet AI creates a shared language for scam events, risk and evidence, helping operators coordinate action while retaining control of their networks and customer policies.”

Making digital trust a shared responsibility

For customers, the intended outcome is fewer scam messages, spoofed calls, robocalls, and Wangiri callbacks reaching their devices. Where a legitimate number is blocked incorrectly, an authorized workflow can reverse the action while retaining the evidence and decision history.

For the wider telecommunications industry, TrustNet AI offers a reusable model for combining adaptive detection, governed action, and cross-industry intelligence sharing. By using common standards without requiring participants to adopt the same platform or share raw customer data, the Catalyst shows how digital trust can become a shared responsibility rather than a problem each operator tackles alone.

Meet the team at Innovate Asia in Bangkok, 17-19 November 2026.