ChatGPT’s Limitations in Reporting Breaking News Raise Trust Concerns

In a digital world racing to embrace artificial intelligence, a significant blind spot has emerged when it comes to breaking news. Early Monday, several major news outlets reported the removal of Venezuelan President Nicolás Maduro, yet leading AI platforms failed to recognize or accurately report this development.

Brian Barrett, Executive Editor at Wired, uncovered this discrepancy after querying multiple AI systems about the situation in Venezuela. His findings revealed troubling inconsistencies in how these platforms processed real-time information.

ChatGPT, OpenAI’s flagship product, categorically denied the event had occurred, stating: “The United States has not invaded Venezuela, and Nicolás Maduro has not been captured.” The system then proceeded to lecture Barrett about potential sources of “confusion,” attributing any such reports to “sensational headlines,” “social media misinformation,” or a misinterpretation of sanctions and rhetoric as military action.

Perplexity AI demonstrated similar limitations, not only rejecting the premise of Barrett’s question but accusing him of falling for misinformation. “The premise of your question is not supported by credible reporting or official records,” the system claimed, before incorrectly asserting that Maduro “remains the Venezuelan president as of late 2025” – a statement containing both factual errors and a bizarre reference to the future.

These failures occurred despite the fact that reputable news organizations including The New York Times, Reuters, and Associated Press had already published verified reports on the situation by the time Barrett submitted his queries.

AI expert Gary Marcus, when consulted by Barrett, explained the fundamental issue: “Pure LLMs are inevitably stuck in the past, tied to when they are trained, and deeply limited in their inherent abilities to reason, search the web, ‘think’ critically, etc.” Marcus emphasized that “the unreliability of LLMs in the face of novelty is one of the core reasons why businesses shouldn’t trust LLMs.”

This incident highlights a structural weakness in large language models (LLMs). While they excel at processing vast amounts of historical text data, they lack the capacity to independently verify current information or maintain accurate real-time awareness of world events. Companies like OpenAI do employ human teams to address these shortcomings, and indeed, when tested later the same day, ChatGPT had apparently been updated to acknowledge the Venezuela situation.

This reactive approach represents the standard operating procedure for generative AI companies: deploying what Marcus describes as “band-aids” to fix specific information gaps as they become apparent. The companies rarely disclose these human interventions, creating an illusion of AI omniscience while masking the significant human effort required behind the scenes.

The implications extend far beyond news reporting. Marcus points to a particularly concerning application: military planning. While Washington policymakers have increasingly framed AI development as a competitive race against China with national security implications, the technology’s fundamental limitations raise serious questions about its battlefield utility.

“If you want to use LLMs to brainstorm or write code, sure,” Marcus notes. “But the idea of using them to plot strategy in rapidly-changing environments like war is laughable.” Military operations require real-time intelligence and rapid adaptation to changing circumstances – precisely the scenarios where current AI systems demonstrate their greatest weaknesses.

The Venezuela reporting failure represents a microcosm of a broader challenge facing generative AI: these systems correlate words from training data but lack “stable, revisable world models.” Without a fundamental understanding of reality that can be updated systematically, AI platforms are condemned to perpetual cycles of error and correction, with human teams rushing to patch each new information gap as it appears.

For news consumers, businesses, and government agencies alike, this incident serves as a sobering reminder that beneath the impressive capabilities of modern AI lies a technology still fundamentally tethered to its training data, struggling to keep pace with the constantly evolving real world.