Inside The Disinformation Industry

Authored by Freddie Sayers via UnHerd.com,

“Our team re-reviewed the domain, the rating will not change as it continues to have anti-LGBTQI+ narratives…

The site authors have been called out for being anti-trans. Kathleen Stock is acknowledged as a ‘prominent gender-critical’ feminist.”

This was part of an email sent to UnHerd at the start of January from an organisation called the Global Disinformation Index. It was their justification, handed down after a series of requests, for placing UnHerd on a so-called “dynamic exclusion list” of publications that supposedly promote “disinformation” and should therefore be boycotted by all advertisers.

They provided examples of the offending content: Kathleen Stock, whose columns are up for a National Press Award this week, Julie Bindel, a lifelong campaigner against violence against women, and Debbie Hayton, who is transgender. Apparently the GDI equates “gender-critical” beliefs, or maintaining that biological sex differences exist, with “disinformation” — despite the fact that those beliefs are specifically protected in British law and held by the majority of the population.

The verdicts of “ratings agencies” such as the GDI, within the complex machinery that serves online ads, are a little-understood mechanism for controlling the media conversation. In UnHerd’s case, the GDI verdict means that we only received between 2% and 6% of the ad revenue normally expected for an audience of our size. Meanwhile, neatly demonstrating the arbitrariness and subjectivity of these judgements, Newsguard, a rival ratings agency, gives UnHerd a 92.5% trust rating, just ahead of the New York Times at 87.5%.

So, what are these “ratings agencies” that could be the difference between life and death for a media company? How does their influence work? And who funds them? The answers are concerning and raise serious questions about the freedom of the press and the viability of a functioning democracy in the internet age.

Disinformation only really became a discussion point in response to the Trump victory in 2016, and was then supercharged during the Covid era: Google Trends data shows that worldwide searches for the term quadrupled between June and December 2016, and had increased by more than 30 times by 2022. In response to the supposed crisis, corporations, technology companies and governments all had to show they were taking some form of action. This created a marketplace for enterprising start-ups and not-for-profits to claim a specialism in detecting disinformation. Today, there are hundreds of organisations who make this claim, providing all sorts of “fact-checking” services, including powerful ratings agencies such as GDI and Newsguard. These companies act as invisible gatekeepers within the vast machinery of online advertising.

How this works is relatively straightforward: in UnHerd’s case, we contract with an advertising agency, which relies on a popular tech platform called “Grapeshot”, founded in the UK and since acquired by Larry Ellison’s Oracle, to automatically select appropriate websites for particular campaigns. Grapeshot in turn automatically uses the “Global Disinformation Index” to provide a feed of data about “brand safety” — and if GDI gives a website a poor score, very few ads will be served.

The Global Disinformation Index was founded in the UK in 2018, with the stated objective of disrupting the business model of online disinformation by starving offending publications of funding. Alongside George Soros’s Open Society Foundation, the GDI receives money from the UK government (via the FCDO), the European Union, the German Foreign Office and a body called Disinfo Cloud, which was created and funded by the US State Department.

Perhaps unsurprisingly, its two founders emerged from the upper echelons of “respectable” society. First, there is Clare Melford, whose biography published by the World Economic Forum states that she had previously “led the transition of the European Council on Foreign Relations from being part of George Soros’s Open Society Foundation to independent status”. She set up the GDI with Daniel Rogers, who worked “in the US intelligence community”, before founding a company called “Terbium Labs” that used AI and machine learning to scour the internet for illicit use of sensitive data and then sold it handsomely to Deloitte.

Together, they have spearheaded a carefully intellectualised definitional creep as to what counts as “disinformation”. Back when it was first set up in 2018, they defined the term on their website as “deliberately false content, designed to deceive”. Within these strict parameters, you can see how it might have appeared useful to have dedicated fact-checkers identifying the most egregious offenders and calling them out. But they have since broadened the definition to encompass anything that deploys an “adversarial narrative” — stories that may be factually true, but pit people against each other by attacking an individual, an institution or “the science”.

GDI founder Clare Melford explained in an interview at the LSE in 2021 how this expanded definition was more “useful”, as it allowed them to go beyond fact-checking to targeting anything on the internet that they deem “harmful” or “divisive”:

“A lot of disinformation is not just whether something is true or false — it escapes from the limits of fact-checking. Something can be factually accurate but still extremely harmful… [GDI] leads you to a more useful definition of disinformation… It’s not saying something is or is not disinformation, but it is saying that content on this site or this particular article is content that is anti-immigrant, content that is anti-women, content that is antisemitic…”

Larger traffic websites are rated using humans, she explains, but most are rated using automated AI. “We actually instantiate our definition of disinformation — the adversarial narrative topics — within the technology,” explains Melford. “Each adversarial narrative is given its own machine-learning classifier, which then allows us to search for content that matches that narrative at scale… misogyny, Islamophobia, anti-Semitism, anti-black content, climate change denial, etc.”

Melford’s team and algorithm are essentially trained to identify and defund any content she finds offensive, not disinformation. Her personal bugbears are somewhat predictable: content supporting the January 6 “insurrections”, the pernicious influence of “white men in Silicon Valley”, and anything that might undermine the global response to the “existential challenge of climate change”.

The difficulty, however, is that most of these issues are highly contentious and require robust, uncensored discussion to find solutions. Challenges to scientific orthodoxy are particularly important, as the multiple failures of the official response to Covid-19 amply demonstrated. Indeed, one of the examples of GDI’s good work that Melford highlighted in her LSE talk was an article on a Spanish website in June 2021 about the Delta variant of Covid-19. “Official data: a third of deaths from the Delta variant in the United Kingdom were among the vaccinated,” reads the headline, next to an advertisement for Chipotle Mexican Grill. “This is clearly untrue,” she said breezily, “and Chipotle has been caught next to this ad unwittingly, and unfortunately for them have funded this highly dangerous disinformation about vaccines”.

This was, however, far from an accurate description. The statistic being reported comes from a June 2021 Public Health England report into Covid variants that sets out the 42 known deaths from the Delta variant from January to June: 23 were unvaccinated, 7 vaccinated with one shot and 12 fully vaccinated. In other words, 29% were fully vaccinated — around a third — and 17% partially vaccinated, making a total of 45% vaccinated. The headline claiming a third were vaccinated, it turns out, was not spreading “dangerous disinformation” at all — if anything, it underplayed the story.

Examples like this are far from rare. The GDI still hosts an uncorrected 2020 blog about the “evolution of the Wuhan lab conspiracy theory” surrounding Covid-19’s origins, which concludes that “cutting off ads to these fringe sites and their outer networks is the first action needed”.

This is despite the fact that Facebook and other tech companies long ago corrected similar policies and conceded that it was a legitimate hypothesis that should never have been censored.

In the US, a number of media organisations have started to take action against GDI’s partisan activism, prompted by a GDI report in 2022 that listed the 10 most dangerous sites in America. To many, it looked simply like a list of the country’s most-read conservative websites. It even included RealClearPolitics, a well-respected news aggregator whose polling numbers are among the most quoted in the country. The “least risk of disinformation” list was, predictably enough, populated by sites with a liberal inclination.

In recent months, a number of American websites have launched legal challenges against GDI’s labelling system, which they claim infringes upon their First Amendment rights. In December, The Daily Wire and The Federalist teamed up with the attorney general of Texas to sue the state department for funding GDI and Newsguard. A separate initiative to prevent the Defense Department from using any advertiser that uses Newsguard, GDI or similar entities has been successful, and is now part of federal law.

But GDI is a British company and, on this side of the Atlantic, the Conservative Government continues to fund it. A written question from MP Philip Davies last year revealed that £2.6 million was given in the period up to last year, and that there is still “frequent contact” between the GDI and the FCDO “Counter Disinformation and Media Development” unit.

Yesterday, I was invited to give evidence to the House of Lords Communication and Digital Committee during which I outlined the extent of the threat to the free media of self-appointed ratings agencies such as the Global Disinformation Index. The reality, as I told Parliament, is that GDI is merely the tip of the iceberg. At a time when the news media is so distrusted and faces a near-broken business model, the role of government should be to prevent, not encourage, and most certainly not fund, consolidations of monopoly power around certain ideological viewpoints.

But this isn’t simply a matter for the media. Both companies and those in the advertising sector also need to act: it cannot be good marketing for brands to target only half the population. Last year, Oracle announced it was cutting ties with GDI on free speech grounds, but as we discovered, it seems they are still collaborating via the Grapeshot plaform: is Larry Ellison aware of this?

At its heart, the disinformation panic is becoming a textbook example of how a “solution” can do more harm than the problem it is designed to address. Educated campaigners such as Clare Melford may think they are doing the world a service, but in fact they are acting as intensifying agents, lending legitimacy to a conspiratorial world view in which governments and corporations are in cahoots to censor political expression. Unless something is done to stop them, they will continue to sow paranoia and distrust — and hasten us towards an increasingly radicalised and divided society.