A new security feature has been unveiled by Meta for WhatsApp. Called Scam Alert, it is an optional AI-powered tool set up to detect potential scam messages directly on a user's device. Scam Alert is being tested with a limited beta audience. It represents one of the most ambitious attempts yet to combine advanced fraud detection with strong privacy protections (and it will keep WhatsApp's end-to-end encryption model intact).

The announcement comes as messaging platforms face increasing pressure from governments, regulators and consumers to combat increasingly complex and sophisticated fraud campaigns. From impersonation scams and fake investment schemes to AI-generated social engineering attacks, criminals are exploiting trusted communications channels at unprecedented scale. Against this backdrop, Meta is betting that on-device artificial intelligence can help identify malicious conversations without requiring access to user messages.

A different approach to scam detection

According to Meta, Scam Alert operates entirely on the user's device. Once enabled, a machine learning model analyses incoming messages from unknown contacts and assesses whether they display characteristics commonly associated with scams. If the system flags a conversation as suspicious, WhatsApp presents a warning visible only to the recipient, who can then decide whether to continue, block the sender or submit a report.

Crucially, Meta says that message content never leaves the device during the detection process. Neither WhatsApp nor Meta receives copies of conversations merely because they have been analysed by the AI model. Reporting remains entirely voluntary and under the user's control.

For years, encrypted messaging platforms have faced criticism that strong privacy protections make it difficult to identify harmful activity. The challenge has been finding ways to detect abuse without creating backdoors or weakening encryption. Scam Alert represents Meta's latest attempt to address that dilemma.

Industry sees a potential new direction

The announcement is likely to attract significant interest across the telecom and messaging ecosystem, where providers continue to invest heavily in anti-fraud technologies.

Many mobile operators and messaging security vendors have traditionally focused on network-level protections, including sender verification, traffic analysis, URL scanning and behavioural monitoring. While these approaches remain important, scammers are increasingly moving to over-the-top messaging applications where traditional telecom controls are less effective.

Industry observers are likely to view WhatsApp's initiative as evidence that fraud prevention is shifting closer to the end user and the device itself. Advances in mobile processing power and smaller AI models now make it possible to run sophisticated classification systems locally, reducing dependence on cloud-based analysis.

The development also aligns with a broader trend towards "privacy-preserving AI", where intelligence is applied without centralising sensitive customer data. This concept is attracting growing attention from regulators and technology providers alike as organisations seek to balance security, privacy and compliance requirements.

A welcome move, but questions remain

While the industry has broadly welcomed stronger protections against scams, questions remain about effectiveness and transparency.

False positives remain a challenge for any automated fraud detection system. Fraud prevention experts note that scam tactics evolve constantly and increasingly rely on social engineering rather than easily identifiable keywords. Success will depend on how well WhatsApp's models adapt to new attack techniques while avoiding unnecessary warnings for legitimate conversations.

Meta has sought to address transparency concerns by committing to publish model information and enabling independent researchers to verify how the system operates. The company is also making use of public transparency ledgers and confidential computing technologies to demonstrate that the system functions as described.

Security researchers will be watching carefully to determine whether these safeguards deliver the level of verifiability promised by Meta and whether the approach can stand up to scrutiny from the broader cybersecurity community.

Implications for the Mobile Ecosystem

The significance of Scam Alert extends beyond WhatsApp itself, as it highlights how the industry's fraud mitigation strategy is evolving from network-centric controls towards a multi-layered approach that combines network intelligence, identity verification and increasingly powerful device-based AI.

Rather than scanning content centrally, Meta's approach demonstrates how machine learning can be deployed directly on consumer devices, potentially offering a model for privacy-preserving scam detection across encrypted communication services.

It also raises broader questions about how mobile platforms, operators, regulators and application providers can collaborate to tackle fraud without undermining user privacy. As governments worldwide intensify efforts to combat digital scams, solutions that can demonstrate both effectiveness and privacy protection are likely to gain increasing regulatory and commercial support.

For the mobile ecosystem, the initiative is notable because it attempts to balance two often competing priorities: protecting users from rapidly evolving fraud threats while maintaining the confidentiality of private communications. As regulators and industry groups continue to scrutinise online scams and digital trust, solutions that combine security, transparency and privacy are likely to receive growing attention.

Whether Scam Alert becomes an industry benchmark remains to be seen. However, Meta's decision to apply AI directly on users' devices rather than in the cloud marks an important milestone in the search for new ways to protect consumers while preserving the foundations of encrypted communications.

Dario Betti is CEO of MEF (Mobile Ecosystem Forum) a global trade body established in 2000 and headquartered in the UK with members across the world. As the independent voice of the mobile ecosystem, MEF focuses on cross-industry best practices, anti-fraud and monetisation. The Forum provides its members with global and cross-sector platforms for networking, collaboration and advancing industry solutions.