Grok, affect software or bulwark towards disinformation? | EUROtoday

On May 25, 2025

Get real time updates directly on you device, subscribe now.

C ‘has turn into an automation. A rumor circulates on X, a surfer wonders and really shortly, the sentence falls: “Grok, is it true?” This reflex, hundreds of individuals adopted it for the reason that look of Grok on the finish of 2024, an AI built-in into the Elon Musk social community. Whether it’s a retouched photograph, a doubtful quote or a declaration attributed to a political determine, the chatbot is commonly known as to the rescue. And, fairly surprisingly, he usually manages to supply a transparent and contextualized response.

Since the repurchase of Twitter by Elon Musk, which has turn into X, the platform has profoundly modified its face. Less moderation, return of former banned accounts, multiplication of conspiratorial speeches: the community is now one of many most important properties of disinformation. In this troubled panorama, Grok is a safeguard-modest, actually, however generally helpful.

Read too The day Elon Musk missed Chatgpt “Grok can be relatively relevant to restore the facts. It also brings an interesting decline on certain viral publications, by adding context or details, “said Tristan Mendès France, a lecturer associated with Paris-Cité University, specialist in digital cultures. The AI has thus refuted unfounded accusations against LFI or dismantled a fictitious quote attributed to Donald Trump calling for Emmanuel Macron.

An AI that corrects … Musk

Ironically, Grok sometimes corrects the conspiratorial figures that Elon Musk himself relays. The AI rejects the idea of ”Woke virus”, defends traditional media, supports diversity policies, as many positions regularly gearous by its creator.

“Chatbots tend to smooth extremes. Their algorithmic logic often produces balanced responses, not very compatible with radical speeches, ”explains to BFM Tristan Mendès France. This doesn’t forestall Grok from being in flip goal of conspiratorial environments, which accuse him of broadcasting progressive biases.

A supply code underneath affect?

However, Grok will not be infallible. And its errors may be critical. The AI notably denied the existence of a picture posted by Donald Trump, when it was very genuine. She additionally relayed conspiratorial theses on the state of affairs in South Africa, or sowed doubts concerning the variety of victims of the Holocaust. XAI spoke of an “unauthorized inner modification” of the system, suggesting that an employee has manipulated the functioning of the AI.

Read too No, the video broadcast by Trump does not show the tombs of South African “1,000 white farmers”Incidents that question, while Grok is increasingly requested to decide between the true of the false in a public space with loss of bearings. Since then, XAI has published part of the Grok code on GitHub. We read clear instructions: never completely trust the institutions, cultivate a constant skepticism. An attractive posture for some, but which flirts dangerously with conspiratorial reflexes.

To uncover

The kangaroo of the day

Answer

We wish to replace you on an incident that occurred with our grok accountability boot on x yesterday.

What Happened:
On could 14 at approximataly 3:15 am pst, an unauthorized modification was made to the grok accountable boot’s immediate on x. This modifications, which direct grok to supply a…

– xai (@xai) May 16, 2025

The AI will not be alone in attempting to oversee info on X. The “community notes” system, launched in 2021, permits customers to right sure content material. But the process, topic to a restricted and generally opaque vote, stays gradual and susceptible to ideological battles. In 2023, Brussels launched an official survey aimed toward X for its presumed breaches of the moderation and transparency obligations supplied for by Digital Services Act. In line of sight, the proliferation of hateful, violent or intentionally false content material. The investigation remains to be in progress.

https://www.lepoint.fr/high-tech-internet/grok-outil-ia-d-influence-ou-rempart-contre-la-desinformation-25-05-2025-2590475_47.php