Safaricom’s bulk SMS filter on hate speech now a global case study

By GATONYE GATHURA

NAIROBI, KENYA: A strategy by Safaricom to filter almost a million bulk messages for hate speech during the March elections is now a global case study on how corporates such as Facebook and Twitter can deal with incitement through their networks.

Despite this, human rights groups feel the company did not do enough to filter hate speech in peer-to-peer generated messages.

During and after the electioneering period, Safaricom had placed tough conditions on bulk SMS, which a new study says had reduced hate speech but unfortunately driven haters to the social media.

In the first of an ongoing series of discussions held in London last month, communication experts advise how governments and corporates can manage public communication without infringing on freedom of expression.

In the discussions expected to feed into the 2014 United Nations Rapporteur on Human Rights report, the enactment of a new media law in Kenya was cited as a serious human rights intrusion.

A resultant 43-page document, Digital Dangers, prepared by the University of Washington and the UK-based Institute for Human Rights and Business (IHRB), details the dilemma faced by Safaricom in filtering out hate speech and at the same time ensuring it did not infringe on the rights of Kenyans to communicate.

POLITICAL MESSAGES

In her presentation, which is available online, Ms Lucy Purdon of IHRB, says within two days to the elections Safaricom had received 68 requests for bulk political messages which translated into 963,762 SMSs.

Ms Purdon had spent the elections week at the Safaricom headquarters in Nairobi.

“At least 18 requests were rejected for various reasons, for example, failing to submit a copy of ID, or specifying who was signing off the message,” says Ms Purdon.

One message concerning land rights was referred to the National Cohesion and Integration Committee as this was considered to be a controversial subject when mentioned in a political message.

The strategy, says Ms Purdon, was largely successful but not foolproof as some undesirable content managed to get through. “In one case a message had gone through without being vetted alleging that a particular candidate was an alcoholic,” she said.

This, Ms Purdon explains, was the mistake of a third party and not Safaricom. In the run up to the elections, Safaricom had put in place regulations requiring the pre-vetting of bulk messages. Bulk messages are those sent from one source to may users such as from Kenya Power to its customers and are managed by third parties for Safaricom.

Such messages had to clearly identify the source while any such messages from outside the country had been banned.

“Some civil society organisations have criticised the guidelines for tackling what they described as ‘low hanging fruit’ and ignoring the wider problem presented by peer-to-peer SMSes,” says Ms Purdon.

According to the Communications Commission of Kenya, four billion SMSes were sent between January and March this year, 94 per cent of them through the Safaricom network. Earlier, Safaricom had said it does not block peer-to-peer messages based on any criteria.

“If Safaricom monitored peer-to-peer SMSes, it ran the risk of breaking privacy laws,” says Ms Purdon.

This, Ms Purdon says in her presentation, would not be dissimilar to issues raised by American whistleblower Edward Snowden on how the US government collects private data on phone calls.

Denied a mass channel to propagate their messages, Digital Dangers says haters seemed to have found a home on the social media. It cites many hate words found on Facebook in days leading and even after elections.

It is estimated that Kenya has two million Facebook users. At a recent meeting on Internet governance in Kenya, Mr Richard Allan, Facebook’s Director of Policy for Europe, Middle East and Africa said the organisation was closely monitoring events during the electioneering period. “If people in Kenya feel that was inaccurate and we could have done more I’d be interested in hearing that feedback,” Allan told the meeting.