Mississippi Publication Accidentally Publishes AI Article

The Mississippi Free Press is a local news site based in Jackson, Mississippi. Backed by the Mississippi Journalism and Education Group, a non-profit organization, the site launched in 2020 and, according to its mission statement, aims to provide “deep, public-interest reporting” on issues facing the region.

However, last week, assistant editor Kevin Edwards published a mea culpa, saying that the publication had unknowingly published an article generated by AI.

The article, published on April 7, was submitted by an outside contributor. Edwards, who is also the editor of the site’s Voices section, said that he only became suspicious when he received an invoice for the article from a name that didn’t match the byline.

Edwards then investigated and found that the people he had corresponded with were either dead or nonexistent. He also found that the headshot submitted with the article was also generated by AI.

He also went back through the other columns he had received and found that many were also from fake authors. Fortunately, the April 7 story was the only one that had been published.

According to Edwards, the organization is “working to take solid steps to raise our level of scrutiny to prevent this from happening again.” However, it’s unclear what those steps will be. In the article, he states that AI detection tools are unreliable and that this is an industry-wide problem.

Ultimately, the story is a reminder that AI-generated content is not just an issue in the classroom or for large publications. It’s also not relegated to spam sites like National Today and Nota’s various sites. It’s a problem that’s reached every kind of writing. Including local sites like the Mississippi Free Press.

A Ubiquitous Problem

To be clear, I don’t want to beat up on the Mississippi Free Press too much. The organization did a lot of things right in this story. They quickly identified the problem, removed the article, and published an editor’s note about the issue.

Given how much AI writing is being published, that is far more than most publications would do or be able to do.

That said, I do wish that Edwards and the publication had been more detailed and transparent about what happened. Which article was it? Who was the author? What specific steps are they going to take to prevent this from happening again? These are omitted from the mea culpa post.

While the acknowledgement is great, it would have been far better with more transparency and information.

But his core point is still valid: this is a ubiquitous problem that they are far from alone in facing.

Simply put, journalism is especially vulnerable here. The rise of the internet has resulted in mass layoffs and consolidation in the news industry. There are simply fewer working journalists than there were 20 years ago, and they are being asked to do more with less.

This has increased reliance on freelancers and outside contributors. This isn’t necessarily a bad thing. As long as contributors are paid fairly, go through the proper vetting process, and their work is properly checked, it can be a good relationship.

But AI has admittedly complicated things. Work from outside contributors is often more difficult to verify, especially when they are a new or one-time contributor.

This is a lesson that academia has been wrestling with for over half a decade. Whether it’s take-home essays or remote testing, assessing the work of remote students has been a growing challenge since the start of the COVID-19 pandemic. AI simply added a new layer onto that problem.

However, this story does provide one potential tool that can at least mitigate the issue for news publications.

The Human Factor

What I found most interesting in this story was not what happened, but how it was discovered.

The journalist, simply put, didn’t exist. Though it took an invoice to unveil the issue, the information had been there all along.

To be clear, different publications have different rules on vetting contributors. If your publication is not checking that a contributor is who they say they are, then adding such a step should be a top priority.

No, this won’t stop humans from abusing AI under their own names. However, it will help weed out some of the worst offenders.

Simply put, it’s bad enough when a human is using AI without proper disclosure. It’s even worse when the human is fake and even the byline is a lie.

Step one should be vetting who the author is. They are the one who will ultimately be held accountable for the work. If they are fake, then there is no accountability to be had at all.

But, while this sounds simple, it’s actually very tedious work. It’s also not something editors are typically trained to do. It’s a simple-sounding solution but it requires resources and time, two things that many news organizations don’t have.

So, even the “easy answer” isn’t easy. But that problem is nothing new to journalism in the digital age.

Bottom Line

While there are things that the Mississippi Free Press could have done better, I have to ultimately recognize that this is a problem that they were not prepared for.

Ten years ago, if I said that fake journalists would submit AI-generated articles to publications, I would have been laughed at. Rightfully so.

However, even then the idea wasn’t that far-fetched. At that time, The Washington Post was experimenting with “automated storytelling” to aid in coverage of the Olympics. Though AI exploded in November 2022 with the public launch of ChatGPT, it’s an issue that’s been on the horizon for a long time.

Still, for news publications, the past few decades have seen them bounce from one crisis to the next. In that context, AI is just another problem that they will have to deal with going forward.

To that end, preparedness is the best defense. Stories like this one should serve as a warning for other publications.

The problem is not just fake articles, it’s fake journalists. That may be the more dangerous problem for the reputation of news organizations in the long run.