India Markets close in 2 hrs 17 mins

WhatsApp: ‘We are not here to give people a megaphone’

Nandagopal Rajan
WhatsApp is using machine learning to crackdown on automated and bulk messaging, which clearly violates its terms of service.

WhatsApp has reiterated that it wants to maintain the "private nature" of its platform and would not like to become a platform for public conversations. The messaging platform has been under the scanner in India after fake news allegedly spread via the platform ended in violent attacks on innocent people.

"WhatsApp was built for private messaging and is designed to share with people you know. It is also built to be data light and provides end-to-end encryption by default in one-on-one as well as group chats," explained Matt Jones of WhatsApp at a media briefing in New Delhi on Wednesday.

The platform, he said, was still used primarily for personal conversations and as much as 90 per cent of chats are sent from one person to another. "Most groups still have under 10 users," he said — WhatsApp groups have a 256-member limit.

"We actively try not to be used for public conversation. So there is a limit now on how many chats you can share a message to," he said giving context on how the limit of five forwards, introduced first in India, has now been extended globally. Incidentally, WhatsApp said one-fourth of the Indian users are not even on groups.

"We are not here to give people a megaphone," Jone reiterated, adding how the platform was using machine learning to crackdown on automated and bulk messaging, which clearly violates its terms of service.

He said that while many were building software and hardware to game the system, these "generate unique behaviour" which can be used to train their systems.

Carl Woog, who heads communication for WhatsApp, said they have "engaged with political parties to express our firm view that we are not a broadcast platform". WhatsApp is releasing a white paper to "ensure there is a clear understanding of how our systems work".

He said they had gone "bottom-up in engaging with parties". Woog said removing end-to-end encryption was not possible and WhatsApp would then "end up being a different product".

Explained: How WhatsApp is using machine learning to fight bulk, automated messaging

WhatsApp said it is using the registration process, messaging behaviour and negative feedback to weed out about 2 million accounts a month out of its 1.5 billion monthly active users now.

Jones said that while the behaviour of those who want to abuse the system was clearly different, it was not simple to isolate as this behaviour is often similar to some patterns noticed with regular users. "Still about 75 per cent users are caught before a single user report is generated against them."

Asked if its actions on automated and bulk messages would affect its business model for companies, Woog clarified to that the legitimate use of its API, where a company was responding to a user-requested information was fine, as it was opt-in by default and based on a WhatsApp reviewed template.

"That is a separate way to sending automated messages and we have a team that watches it closely."