In 2018 Twitter CEO Jack Dorsey acknowledged that employees at his company have a “left-leaning bias” – but insisted that it does not affect how Twitter makes decisions on content on its platform.
The first part of his statement was seized on by right-leaning commentators, in the USA and abroad; the second part was derided and dismissed. Donald Trump and many of his supporters have long suggested liberal bias at the California-based social media companies and that they have crossed the line into censorship. In the context of an election year, a collision was inevitable – and the consequences may be long-lasting for freedom of speech everywhere. While the election is over, President Trump is not yet gone – and change may be coming even under a Biden administration.
When, on 27 May 2020, Twitter made the (prescient) decision to fact-check as “unsubstantiated” one of the then-President’s tweets alleging widespread fraud in mail-in voting, it was a shot heard around the world. Donald Trump responded characteristically – with an Executive Order threatening to narrow the legal protections conferred on providers of an interactive computer service by section 230 of the US 1996 Communications Decency Act.
Section 230 is how, in conservative eyes, providers get away with unconstitutional censorship in modern America – through abusing moderation rights. Our previous Trump v Twitter piece explained how s.230 gives both a shield and a sword to online providers like Twitter and Facebook, providing not only (1) that no provider shall be treated as the publisher of information provided by another (thereby protecting providers, as supposed mere platforms, from legal proceedings regarding what has been said by users on those platforms), but also (2) that where a provider does, in good faith, moderate what is published by American users, such moderation is not an infringement of those Americans’ right to freedom of expression under the First Amendment of the US Constitution.
Trump’s 27 May Order threatened to upend this. If providers want to keep these two legal immunities, his Order said, they would have to avoid “editorial” actions inconsistent with being a mere platform. Our previous article suggested this was an essentially Damoclean threat, warning the social media platforms off doing anything which might drive a wedge between the President and his base in this election year, rather than something which would actually be implemented. if – as he no doubt considered – Twitter called his bluff, this would still enable Mr Trump to rail against the “liberal left” conspiracy.
Twitter chose not to get the message. Within hours of the Order, Twitter hid one of the President’s tweets entirely – for “glorifying violence“. Both Twitter and Facebook removed his “tribute video” to George Floyd for breach of copyright. Coronavirus forced a further escalation: in October Twitter removed a tweet from Trump adviser Scott Atlas suggesting the wearing of face masks as a precaution against the spread of Covid-19 were ineffective or harmful: the Tweet was adjudged to be “sharing false or misleading content which could lead to harm”. Twitter also acted to suppress posts referencing an anti-Biden New York Post article, supposedly because of doubt over the veracity of the underlying material. The Republican-dominated Senate duly subpoenaed Jack Dorsey and Mark Zuckerberg to testify on their companies’ activity – focusing on the response to the New York Post article.
As the November 2020 election came around… and then dragged on… Twitter and the other platforms moved into political monitoring and real-time fact checking, hiding or flagging tweets which called the election or denounced the results. Moderation response times on Twitter dropped to thirty minutes or less. By 15 November even the language of studied neutrality was strained to breaking point:
“I WON THE ELECTION!”
“Official sources called this election differently”
The implications of the social media platforms’ defiance of Trump’s Executive Order are still reverberating. In refusing to be cowed by the Order, Twitter and other platforms may have been gambling that legislation to implement the Order would never make it to law. Trump is still trying to deliver change – on 1 December he threatened (via Twitter, of course) to veto a US$740 billion defence bill unless repeal of the “very dangerous & unfair” s.230 was tacked on. However, commentators believe he is unlikely to make good the threat (that annual bill hasn’t been vetoed in six decades) and eyes are now turning to what does appear to be an impending Biden administration. The question is whether, even under Biden, s.230 will remain in force in its current form.
In this, the 17 November Senate hearing for Jack Dorsey and Mark Zuckerberg’s subpoenas was instructive. We saw Republican and Democratic lawmakers’ shared antipathy to s.230 in its current form – and their equally absolute disagreement as to how, and indeed why, it should be replaced. Many Republicans said it too easily allows unconstitutional constraint of free speech – but many Democrats were also hostile, saying s.230 does not do enough to prevent harmful publications.
The first question under discussion in the Senate was of course whether the social media platforms did, and do, too much to moderate content. In s.230 terms, were they acting in good faith as moderators? Many Republicans say not (citing Twitter’s energetic clampdown on retweets of the anti-Biden New York Post article as particular evidence of uneven, anti-conservative, moderation) and that repeal/the reform proposed by the 29 May Executive Order is necessary to prevent constraint of free speech, contrary to the First Amendment. Many Democrats say in response that the providers were absolutely acting in good faith, because what was being published was simply untrue, and dangerous.
Indeed, the second, diametrically opposed, question under discussion, predominantly on the Democratic side, is whether the social media platforms did, and do, enough to weed out objectionable content, including disinformation. If they do not, should s.230 immunities be retained – or are they doing more harm than good? In 2019 Democrats were (amongst other matters) outraged by the ability of a terrorist to live-stream the first of his mass shootings in Christchurch, New Zealand on Facebook; more recently they have been (understandably) concerned by a proliferation of American groups using social media to organise and to seek to impugn both the US election process and medical advice on the coronavirus response. If platforms will not take faster and wider action against this kind of content, strengthening moderation, then a Democratic revocation and replacement of s.230 will certainly be on the agenda. Many Republicans, of course, say that President Trump and others have a right to speak absolutely freely – including to publish “alternative facts”.
This issue really matters: recent polls suggest that (notwithstanding the absence of any real evidence so far) around 68% of Republican voters said they were concerned that the US election was “rigged”. How did this view enter the popular consciousness? Through the social media of President Trump and others, relentlessly pumping out the message of a rigged election, before and after the vote – an allegation which had to be then reported in the mainstream press even as it was dissected. Recent analysis suggests that of Trump’s 550 Tweets in the three weeks following the election, about three-quarters attempted to undermine the election results. In a social media cacophony, the assumption that the US courts’ decisions on these allegations will be the last word can no longer be relied on.
President-elect Biden is on record as supporting revoking s.230 protection. But ultimately the question may be decided in the Senate – and may therefore, like so much else, be ultimately settled in Georgia. If President Biden does not gain control of the Senate through the two runoff votes there, on 5 January 2021, then he would need bipartisan support to arrange any revocation and replacement of s.230 – and, if 17 November’s Zuckerberg/Dorsey Senate hearing is any guide, bipartisan support may be very hard to come by.
The Global Impact
The first question for those of us not in the USA is whether platforms like Twitter, used globally, will extend their fact-checking to other elections. Last year the University of Oxford’s Internet Institute found “organised social media manipulation campaigns” now prevalent in 70 countries around the world – and that number is not likely to shrink. Will other elections – India, or Brazil, or Nigeria – be handled with the same real-time care and attention? If not, why not? Was this particularly fraught election a special case, deserving special moderation, not to be repeated? Online platforms may find it hard to justify a retreat from the measures they adopted for 2020.
The question applies equally in the UK. Could we see a Twitter employee sitting in California, or Ireland, deciding whether the British Prime Minister is telling the truth?
The second question is whether roll-back of s.230 and/or increased voluntary moderation under a Biden administration will change how providers handle foreign content. Even if he doesn’t win the Senate, the President-elect has a reputation for conciliation: a Biden administration will probably use subtler methods than the 29 May Executive Order (starting with engagement with the tech companies) to encourage greater moderation. If the threat of s.230 revocation is used to leverage reform, that will avoid the actual need for revocation. A voluntary shift will still be an uneasy (and, for the companies involved, expensive) compromise, but it will be less violent – and therefore it may take longer to influence their overall philosophy and operations outside the USA.
What appears (at time of writing, at least) to be Donald Trump’s slow fall may put all these issues on the back-burner for a while – but it will not be long before they return to influence us all.
Oliver Cox is a Senior Associate at Carter-Ruck, specialising in media law matters and commercial dispute resolution.