"In the likely absence of state intervention, and inaction on the part of the social network providers, users may have no choice than resorting to take matters into their own hands" thinks Chris Blackhurst, journalist at The Independent.
"One such is the British company Arwen, which is just celebrating its second birthday. Founded by Matthew McGrory, who used to run IT for Fulham Football Club and Brands Hatch, before going into app development, Arwen uses AI to continually scan its subscribers’ Twitter, Facebook and Instagram feeds. Artificial intelligence searches every comment they receive in real time, looking for 24 different types of abusive and unwanted content across 29 languages.
The comments are classified as “safe”, “suspect” or “severe”. The latter is hidden from view in under a second. “Suspect” comments are examined closely and “safe” speaks for itself. Repeat offenders can be blocked and if needs be, reported to the network operator and the police.
The subscriber gives permission to collect their data, 24 hours a day, 365 days a year, and then Arwen goes to work, says McGrory. They’re looking for hateful language but also unsolicited material that may cause offence, like bots popping up in a message promoting porn or cryptocurrencies. “Brands, corporates, celebrities, individuals, they don’t want the association with that sort of stuff, they want it removed.”
The AI can also earmark escalating threats. “We’re offering free trials to MPs at the moment, to help them locate bad actors,” he says.
It’s clever technology able to deal with ambiguities. “It can flag the particular use of a word and yet pass the same word in another context. So, for instance, there can be mention of a black-tie event, which is harmless, but the same person can receive a message calling them a ‘black beep beep’ which is clearly offensive.”
At present, they’re claiming a 92 per cent success rate, but they expect that to rise to 95 per cent and higher still, meaning virtually all the material the client deems offensive is removed.
For individuals with up to 200,000 followers, Arwen charges £30 a month; for those with more (one customer is Lewis Hamilton with 43m followers) the fee climbs to a maximum of £1,000 per month.
The firm is still in its infancy, but the numbers are rising rapidly. So far, Arwen has signed up 140 social media accounts, including those of around 50 corporate brands. They include Premiership football clubs. In all, McGrory claims “we’re protecting 75.1 million people from toxicity, spam and unwanted content every minute of every day”.
It’s a pity that this type of service is only available to those who can pay. “The government is indicating it won’t address the toxic abuse side of things in the Online Safety Bill,” says McGrory. “The door is left open for us.”