In the age of social media and instant information, fake news and online disinformation have become urgent threats. As bad actors exploit platforms to spread propaganda and sow discord, the tech world is turning to an unlikely ally in the fight for truth: artificial intelligence.
Enter large language models (LLMs) — powerful AI systems trained on vast datasets capable of generating human-like text with uncanny fluency. Led by OpenAI's ChatGPT, these models are revolutionizing industries from customer service to creative writing. But could they also help stop fake news in its tracks?
LLMs present both promise and peril. On the one hand, they can be trained on legitimate news to learn patterns that distinguish credible reporting from deceptive content. As such, they can serve as automated fact-checkers, scanning online content and flagging suspicious claims.
Promising initiatives already exist. MIT and IBM researchers developed the Giant Language Model Test Room (GLTR), which uses an earlier GPT model to detect AI-generated writing by identifying unusual word patterns and phrasing — tools useful for spotting fake articles and forged documents.
However, the same technology can also be weaponized. With the ability to generate realistic text at scale, LLMs could be exploited to mass-produce fake news, social media posts, and even entire disinformation campaigns.
"The same technology that can be harnessed to spot fake news could just as easily be weaponized to create it."
The concern is real: a technological arms race is underway. Microsoft, an investor in OpenAI, has integrated ChatGPT into its products. Meanwhile, Google and Meta are building rival language models.
As these systems become more sophisticated, so do the risks. Experts fear we may soon face an era of “deepfakes for text” — AI-generated disinformation indistinguishable from human-written content.
Policymakers are scrambling to respond. In the U.S., hearings have been held on AI and disinformation, but legislation lags behind. In Europe, the proposed AI Act has sparked debate for being both sweeping and vague.
Meanwhile, tech companies operate with little oversight. Though OpenAI prohibits impersonation or explicit content generation, critics argue existing safeguards are insufficient.
"The breakneck speed of AI development has left governments flat-footed, with few clear guidelines on how to police the use of language models."
What can be done?
"As language models become more sophisticated, it will be increasingly difficult for the average person to spot fake content generated by AI."
LLMs like ChatGPT offer powerful tools in the fight against disinformation—but they also raise the stakes. As AI technology accelerates, policymakers, technologists, and citizens alike must work together to build systems that support truth, not undermine it.
The decisions we make today will shape our information ecosystem for decades to come.
Stay up to date with expert insights on AI and media integrity by subscribing to our weekly newsletter. Don’t miss the latest developments shaping the future of truth online.
.svg){kind=small}