OpenAI VP Retracts Tweet Containing Inaccurate Claims About GPT-5’s Mathematical Abilities

In a digital era where artificial intelligence breakthroughs often capture headlines, a recent misstep by an OpenAI executive has highlighted the fine line between technological enthusiasm and factual accuracy.

Last week, Kevin Weil, a vice president at OpenAI, published a statement on social media claiming that the company’s upcoming GPT-5 language model had “found solutions to 10 previously unsolved Erdős problems and made progress on 11 others.” These problems, named after legendary Hungarian mathematician Paul Erdős, represent some of mathematics’ most challenging unsolved conjectures.

The claim quickly attracted attention from the scientific community, but not for the reasons OpenAI might have hoped. Thomas Bloom, a mathematician at the University of Manchester who maintains erdosproblems.com, publicly challenged Weil’s assertion, calling it a “dramatic misrepresentation” of the AI’s capabilities.

According to Bloom, what GPT-5 had actually done was far less revolutionary – the system had merely retrieved existing solutions that were already documented in mathematical literature, rather than developing novel approaches to unsolved problems. The distinction is crucial in a field where original discovery carries significant weight.

As scrutiny mounted, Weil deleted the tweet, but not before it had sparked widespread discussion across social media platforms and specialist forums. The incident was reported in detail by technology publication Futurism, among others, drawing further attention to the gap between the claim and reality.

This episode comes amid OpenAI’s broader promotional efforts for GPT-5, which the company has described as possessing “PhD-level” intelligence. Industry experts note that truly solving previously unresolved Erdős problems would represent an extraordinary advancement in artificial reasoning – a capability that would transcend the pattern-matching techniques that underpin current language models.

The AI research community’s reaction was swift and pointed. Demis Hassabis, CEO of Google’s DeepMind, appeared to reference the controversy in comments emphasizing the need for scientific rigor in AI claims. On platforms like Reddit and X (formerly Twitter), technical discussions dissected the validity of OpenAI’s assertions and questioned the company’s internal verification processes.

This isn’t the first time AI companies have faced criticism for overstating capabilities. The competitive landscape, coupled with immense investment pressures, creates incentives for amplifying successes. OpenAI, valued at tens of billions of dollars, operates in an environment where demonstrating advancements directly impacts financial prospects and public perception.

When contacted about the controversy, an OpenAI spokesperson acknowledged the error but characterized it as stemming from enthusiasm rather than intentional misleading, according to Futurism’s reporting. This response has done little to quell concerns about how technological achievements are communicated to the public.

The incident also highlights evolving questions about AI’s role in scientific research. While large language models can serve valuable functions in literature review and hypothesis generation, conflating information retrieval with genuine discovery undermines both technological credibility and scientific integrity.

For the broader AI industry, this controversy emerges at a sensitive time. Regulatory bodies worldwide are increasingly scrutinizing AI claims and capabilities, with particular attention to transparency and accuracy in corporate communications. Critics, including OpenAI co-founder Elon Musk, have pointed to incidents like this as evidence of prioritizing marketing over substantive progress.

The ramifications extend beyond temporary embarrassment for one executive. As AI systems integrate more deeply into critical sectors like healthcare, finance, and public infrastructure, the stakes for accurate representation of capabilities grow higher. Public trust depends on responsible communication that distinguishes between speculation and verified capabilities.

For a field that prides itself on data-driven precision, maintaining communication integrity represents a fundamental challenge. This episode may ultimately serve as a cautionary tale for AI developers and communicators, reinforcing the importance of verification before proclamation – especially when the claims involve breakthroughs in fields as rigorous as mathematics.