How to protect the media from misinformation?

How is information organized?  

The contemporary era marks a split in the journalistic model as we know it, as Franck Rebillard explains in his book From information processing to reprocessing. The history of the written press in its modern form goes back to the 19th century. The industrial era, marked by the intensification of trade, the development of rotary presses, and the progress of education, offered a favorable environment for the press, which then experienced the beginning of its golden age, until the second half of the 20th century when radio and then television brought new audio and then video formats. 

Between 1803 and 1870, the circulation of the Paris daily press increased from 36,000 to one million copies, but access to international information was complex and costly due to telegraph rates of between 100 and 700 francs per kilometer, or several hundred euros. 

To overcome this, three young press agencies, Havas, founded in Paris in 1835, Reuters in London in 1851, and the Agence Continentale, established in Germany in 1849, formed a cartel from 1859 onwards, centralizing international information while establishing a monopoly in their respective countries. The arrival of the first telegraph cable from America in 1865 forced them to open this cartel to the Associated Press, created in 1846 in New York. This association made it possible to reduce costs while at the same time advertising emerged as an indispensable source of income. 

Thus, the press agencies settled into their role, which was to collect, verify and then distribute information to the various press titles. To do this, they rely on correspondents all over the world, incomparable institutional networks, and large editorial offices. The press agencies also centralize the information published in the media after having carried out a verification process. Agence France Presse (AFP), which was created from the split of Havas into two activities, the second being the communication agency of the same name, has 2,400 employees, including 1,700 journalists working in 151 countries. These statistics are similar to those of its British competitor, Reuters, which publishes over 2 million dispatches, nearly 130,000 video reports, and a hundred or so investigative reports distributed to more than 2,000 clients worldwide each year. Together with Associated Press, they are the three leading agencies in the world. While these are the most prominent, there are general news agencies in almost every country, working alongside agencies specializing in a particular field, such as economics, art, and defense, or in a particular mode of communication, such as photography or video content production. 

All the content produced is then distributed to the agencies' media clients, who either republish it as is or add to it. Dispatches that are republished without significant modification can generally be identified on media websites by the words "with AFP" or "with Associated Press".

Media outlets operate similarly, with the triptych of sources, outlets, and distribution media. Raw information is gleaned by investigation, from sources, via private and public documents, or more simply from press releases issued by institutions, organizations, and companies. 

The advent of digital technology has further changed access to information as journalists have lost their monopoly. Everyone is now able to disseminate information to a wide audience. Above all, the media have changed their temporal paradigm. In the days of the written press, information was part of a precise temporality. Editorial offices prepared editions that had to be completed by the end of the day and then went to print before being distributed throughout the country. 

Today, information is available continuously. With the advent of television, correspondents have seen their missions evolve by going live to events. This has reduced the time needed to prepare stories. In the digital age, this tempo has accelerated even more since anyone can report on an event and it is not uncommon to see amateur images illustrating stories. This raises the issue of partial illustration, which can alter the veracity of the facts and lead to the dissemination of false information. There is also a risk of publishing, in the rush for speed, amateur videos, and images that correspond to another event, leading to a form of disinformation. 

How can false information be spread in the media? 

In November 2016, the Bloomberg affair hit the headlines. The financial news agency relayed a false press release allegedly published by the Vinci group and reported financial malpractice to the tune of 2.4 billion euros. This document directly implicated the financial director, who was announced as having been dismissed by the group. We could read in this press release: "Vinci is launching a review of its consolidated accounts for 2015 and the first half of 2016 [...] The results of an internal audit conducted by the Vinci group have indeed revealed that certain irregular transfers had been made from operating expenses to the balance sheet, outside of all recognized accounting principles [...] The amount of these transfers would amount to €2,490 million [...] The company has dismissed Christian Labeyrie, deputy general manager and financial director of Vinci."

The forgers had produced a press release that was faithful to the graphic charter, the only difference being the telephone number of Vinci's press officer. At 4.05 pm, this false document was sent to the press agencies, including Bloomberg, followed by almost all the financial players. Only one minute later, the information was relayed by the agency and at 4.15 pm, Vinci's share price had collapsed to 49.93 euros from 61.56 euros nine minutes earlier. At 4.44 pm, journalists received a denial from the group, which was equally false, but once again true to the group's charter. Vinci's real press release did not arrive until 5.02 pm. Between the two false releases, Vinci's capitalization had lost 7 billion euros (-19%) and at the close of the CAC40, it was still down -3.76% while the stock market was up slightly by +0.49%.

This case, in which Bloomberg was sentenced to €5 million in the first instance by the AMF (Autorité des Marchés Financiers), the french stock market regulator, a sentence reduced to €3 million in the second instance, reveals how false information can be relayed under time pressure. The AMF's decision shows that the "wrongdoers" were two members of Bloomberg's "Speed Desk" service, which "consists of the publication of real-time financial information extracted from press releases or other sources and relayed in the form of flashes or alerts. As Marianne explained, this practice, which consists of relaying a press release as quickly as possible, relegates the journalist to the role of a "transmission belt" and the veracity of the press release is never questioned. 

Without deliberate deception, the dissemination of false information can be the result of human error or a combination of circumstances, which reveal flaws in the verification process.

On 28 February 2015 at 2.28 pm, AFP announced the death of Martin Bouygues, Chairman, and CEO of the eponymous group, in a dispatch: "Martin Bouygues died on Saturday morning at his home in the Orne region (town hall)." The media immediately picked up this information and widely disseminated it in just a few minutes. At 2.42 pm, the agency published a second dispatch intended to clarify the previous one. "The industrialist Martin Bouygues died on Saturday morning at the age of 62 in his home at La Roche Mabile, near Alençon (Orne), according to the mayor of the neighboring commune of Saint-Denis-sur-Sarthon. "I can only confirm his death," mayor Michel Julien told AFP, refusing to specify the consequences (sic) of the industrialist's death. At 2:46 pm, the excitement accelerated with the release of the information by the British agency Reuters, citing France Info, which had previously relayed the AFP text. At 2.55 pm, after also announcing the death of the company director, the channel LCI, a subsidiary of the TF1 group owned by the Bouygues group, announced that its CEO was alive and well. Catherine Nayl, deputy director general of information for the TF1 group, said: "I categorically deny this information because I spoke to Martin Bouygues myself on the phone about ten minutes ago, that he is fine, that he is completely surprised by this announcement, that he cannot explain it. So, I deny this information and I tell you that Martin Bouygues is fine. 

So how could this mistake have happened? It would appear to be a three-act incident that began with a report, of unknown origin, that Martin Bouygues had died in his home in the Orne region of France, which reached the AFP newsroom. The Rennes office immediately contacted Michel Julien, the mayor of Saint-Denis-sur-Sarthon, a neighboring commune of La Roche-Mabile, where Martin Bouygues had died. There was then confusion over the surname. When asked by the journalist, the mayor referred to a Mr. Martin. He declared on BFM TV: "The journalist said to me 'Mr. Martin is dead'; I said 'yes' [...] I am very surprised, I didn't know we were talking about Martin Bouygues.

Finally, the Rennes bureau sent its dispatch to Paris, which was published without further confirmation before 4 pm, and AFP sent a message to its clients: "Please disregard the whole series of dispatches (alert, urgent, LEAD) concerning the erroneous announcement of the death of Martin Bouygues. 

Afterward, Rémi Tomaszewski, the agency's managing director, deplored the situation, "we tried to confirm and not to verify". Especially since Mr. Bouygues has no ties to the Orne region. The damage to the agency's image and credibility was significant in the immediate aftermath, but had no financial consequences, as the agency's president Emmanuel Hoog points out: "None of AFP's clients canceled their subscriptions, and none of them blamed us. It is at times like this that big organizations show their strength. The three big news agencies, Associated Press, Reuters, and AFP, have always been able to evolve.

In these two examples, among the most striking of the last decade, the desire to react quickly was the main reason for the failure. The agencies in question have daily demonstrated their ability to deliver a large volume of reliable information. However, competition with new information channels that can deliver information to the public on a massive scale and the challenges of building public trust in the media are new challenges. 

According to the Digital News Report 2022 conducted by Reuters Institute, European citizens' trust in the media continues to erode. For example, British trust in the BBC has fallen steadily over the past five years, from 75% in 2018 to 55% in 2022. In France, the overall level of trust in the media in 2022 is 29%, 9 points lower than in 2015, and Le Monde, the main national daily newspaper, has only 49% trust.

These challenges oblige the players in the news industry - agencies, newspapers, television, and radio stations - to continue to adapt and adopt new methods capable of guaranteeing the authenticity of the information. Lastly, the emergence of artificial intelligence makes it possible to support journalistic action by facilitating these tedious tasks and avoiding human errors.

How can Buster.Ai be used to verify information in the media?

Buster.Ai is a French company founded in 2019, specializing in the fight against disinformation thanks to advanced artificial intelligence. As Julien Mardas, its founder and CEO, confided on the air of BFM Business, "Buster.Ai first and foremost are humans who work in research to teach a machine to read and by teaching a machine to read, we teach it to understand the information, to represent it in its symbolism and to compare it to millions of sources in real-time to detect whether there is disinformation or not, and above all, what its impacts will be." 

The media can therefore sift and filter crucial information to validate or invalidate hypotheses, which allows them to ensure the veracity of the facts before publishing information.

How do you fact-check a massive flow of information? 

The media is fed by a continuous flow of information from a wide variety of sources. Whether it is news agency dispatches, company press releases, interviews, or elements from their own sources, the media are very rigorous in the choice and construction of their stories. However, due to the massive flows, manual processing is extremely time-consuming and risky when it comes to checking the data. Thanks to the Buster.Ai API, newsrooms benefit from a tool that can handle a large flow of queries while being reliable and efficient in its verdicts, enriched by institutional, academic, and media sources, thus facilitating the research work of journalists. 

How could Buster.Ai have prevented the spread of fake news in the cases mentioned in this article?

Let's first look at the Vinci Bloomberg case. Deploying Buster.Ai's API in a news agency's editorial office allows for live verification of a massive news flow without altering processing times to allow for quick publication while being pre-checked. In this particular case, the news agencies would have benefited from a verdict that would have refuted the news, beyond the usual prudence in the face of information. Indeed, the tool, which relies on millions of sources, including the institutions, would have shown that no such press release had been issued by the group and that the latest communications did not mention any wrongdoing. 

As for the "fault" committed in the AFP editorial process, the API proposed by Buster.Ai could have enlightened the journalist on the businessman's personal and family ties or on his latest news, which could have informed him about his agenda.

Buster.Ai, the ideal solution for fact-checkers?

More generally, Buster.Ai is becoming a key player for the media and provides solutions to the challenges they face by offering artificial intelligence models capable of reading texts, constantly enriching their models, and analyzing thousands of sources to deliver not only extracts from relevant sources but also a "Supported" or "Refuted" verdict. This solution, which is accessible via API, is reliable in its verdicts and fast in its execution, making it possible to understand the spread of information, whether true or false.  

Contact Buster.Ai if you would like to know more about how we can help protect you from disinformation.