Elon Musk promised to defeat Twitter’s spam bots. He’s failing
- Insight Partners, a New York-based venture capital firm, warned it was witnessing a “great reset in tech” after valuations in the sector slumped.
- Saudi Arabia spent almost $8 billion building stakes in gaming companies.
- Paul McCartney said he used artificial intelligence to create “the final Beatles record.”
When he announced his bid to buy Twitter in April 2022, Elon Musk made a promise: “If our bid succeeds, we will defeat the spam bots or die trying.” Just over a year later, Twitter’s bot problem might – if anything – be getting worse.
So what? Bots – or automated social media accounts – make up as much as 40 per cent of all web traffic and are a key driver of misinformation and fraud. On social media platforms, they have the power to influence elections, scam users and annoy advertisers, who lose billions of dollars each year on ad fraud.
One of Musk’s most important changes as owner has been to revamp Twitter Blue – Twitter’s subscription service – and tie it to user verification. By asking users to pay $8-a-month for a blue tick, a process which requires a valid credit card, the SpaceX founder argues that bots will no longer be able to operate at scale.
Has it worked? Not yet. Well-respected cybersecurity firm Cheq found this week that the percentage of ad traffic driven by bots was roughly 12 per cent in the first quarter of 2023 – around the same as it was a year ago. Its findings, first reported in the WSJ, were based on 2,000 real-time tests. Researchers accept that it’s hard to determine exactly how many bots exist on Twitter, as the site operates a technological walled-garden.
What’s going wrong? For his plan to work, Musk – and Twitter’s new CEO, Linda Yaccarino – need millions of real users to sign up to Twitter Blue. Yet as of February, only about 300,000 people have reportedly done so. In addition, scammers don’t need to buy large numbers of paid-for accounts as a single bot can have an outsized effect.
Some believe that Musk’s changes have added to the problem. In the weeks after Twitter Blue launched, blue-ticked accounts impersonating Tony Blair and George Bush went viral, while US pharma company Eli Lilly had to apologise after a “verified” account promised that “insulin is free now.”
You ain’t seen nothing yet. Musk argues that paid-for verification is the only way to address “advanced AI bot swarms taking over”. He might be right. Large language models like Chat GPT and Facebook’s LLaMA have the potential to run hundreds of thousands of fake accounts in parallel – and have already started causing problems.
If you search for the phrase “as an AI language model”, you can see examples of Chat-GPT-driven bots operating today. One AI-driven network identified in April had 60,000 accounts. Last month, multiple accounts tweeted a bogus image of an explosion at the Pentagon – likely made with generative AI.
Things are only going to get worse from here.
