Gamers Trick AI-Generated News Site Into Posting Fake Article

Gamers have tricked a website into publishing an apparently AI-generated news article based on their comments about a new World of Warcraft feature that doesn’t exist.

The incident comes as experts and activist groups warn of the potential of so-called generative artificial intelligence – which is capable of creating human-like text or images – to supercharge online fraud and misinformation.

Gamers in Reddit’s World of Warcraft subreddit took action to highlight the use of news bots after noticing that AI-generated articles were being produced that simply summarized the contents of popular threads on the social media site, apparently with little or no human involvement.

They pitched together on a thread titled “I’m so excited they finally introduced Glorbo!!!” in which they excitedly discussed the supposedly long-awaited “Glorbo” World of Warcraft feature – one that is not real.

Fake news

A gaming-oriented app maker and social media site called zleague.gg soon took the bait and published an article delivering a summary of the thread, using a clear and comprehensible style typical of generative AI, but with no regard for the actual content of what was being discussed.

“Honestly, this new feature makes me so happy! I just really want some major bot-operated news websites to publish an article about this,” the article quotes the original Reddit user as saying.

The article – which appeared under a human byline – goes on to summarise posters’ increasingly bizarre comments about the “Glorbo” feature, such as players’ differing opinions about “mandatory item Klikclac”, the “rumoured new location for the Halfhill Market” and “the inclusion of Klaxxi as a playable race”.

It dutifully makes note of one user’s comments that the feature is the best addition to World of Warcraft since “the quest to depose Quackion, the Aspect of Ducks” – another nonexistent feature.

AI content

One of Warcraft’s senior developers joined in on Twitter, linking to the article with the comment, “Feels soooooo good to be able to talk about Glorbo finally, I remember my first day at Blizzard we were just starting to work on implementation, and that was almost 15 years ago! Excellent reporting to track this down.”

Zleague later removed the article, which is archived online, but continues to post dozens of reports per day under the same byline.

While Zleague is primarily an app developer, major content producers have begun investigating the use of generative AI, in a trend some have found concerning.

One of the highest-profile examples is tech and consumer site CNET, which in January was found to have published AI-generated articles with extensive structural and phrasing similarities to human-written pieces from competitors such as Forbes – in addition to a number of factual errors.

Legislation

CNET paused its use of AI-written articles at the time, and in June published an AI policy pledging  transparency.

Such concerns have led governments around the world to begin developing AI laws, including the EU with its upcoming AI Act.

Meanwhile the US’ Biden administration last week promoted a voluntary pledge by some of the world’s biggest AI companies to work on AI safeguards such as watermarks, intended to make it easier to spot AI-generated content.