MIT Technology Review Warns AI Hype Is Outrunning Expert Scrutiny

The sequence began at the end of April, when Anthropic claimed that its model Claude Mythos was better at finding software vulnerabilities than most security experts, according to the article. It then cites the OpenAI–Hugging Face hacking incident, after which Anthropic disclosed similar incidents involving its models and Meta did so reluctantly. Later, Anthropic claimed one of its models had made a mathematical breakthrough, and soon OpenAI claimed a mathematical breakthrough of its own. Most recently, Anthropic engineer Jacob Coxon went viral announcing his departure from the company, saying that it and OpenAI are 'racing straight towards self-improving superintelligence and gambling with our lives.'

MIT Technology Review says such episodes are often covered with little skepticism, with media repeating the companies’ anthropomorphizing framings. Those framings are designed to portray their software as powerful and as incipient 'artificial general intelligence.' The article asks whether the public is witnessing civilization-changing technological breakthroughs or marketing.

On the hacking incidents, cybersecurity experts say the story is more about OpenAI’s negligence and failure to adopt basic, established security practices than about models gone rogue or 'AI agents creating civilizations,' according to the article. On the mathematical results, mathematicians who were initially stunned by OpenAI’s press release saying its latest chatbot, Astra, solved problems that 'have been open and seen no progress on the main result for at least a decade' later realized the results were not as novel as they first appeared. They have since accused the company of research misconduct and plagiarism and reiterated that Astra did not make a 'profound intellectual leap.'

Weeks later, OpenAI claimed its own mathematical breakthrough. Two days before, Tristan Buckmaster, a math professor at New York University’s Courant Institute, published a statement suggesting OpenAI had stolen other people’s work and improperly attributed it, the article says.

The article argues that claims of incipient, dangerous superintelligence are not based in good scientific or engineering practice. It describes them as narratives rooted in ideologies of transhumanism, eugenics and wishful thinking about imagined future digital humans. It also examines why computer programming and mathematics receive so much attention as fields for large language models and related technology. It says these fields are often elevated as the pinnacle of human intellectual achievement, and they involve problems whose answers, once suggested, can be verified. The first trait helps sell the idea that companies are building 'everything machines,' while the second makes math and coding problems easier to tune systems for, because output can be evaluated without paying data workers to review and annotate each item.

Mathematicians have warned against corporations using their field this way. A statement signed by hundreds of them says there is 'currently a strong commercial incentive on the part of the technology industry to overstate the capabilities of their products' and asks policymakers to 'consult with experts, including mathematicians, in forming policy decisions rather than relying on press releases or popular reporting of mathematical results.' MIT Technology Review says it echoes that call and adds that the illusion of speed and urgency promoted by tech companies is also a ploy to misdirect policymakers and the public. It points to Senator Bernie Sanders’s proposed legislation to prevent the development of 'artificial superintelligence' as well-meaning but ultimately misguided.

The article says describing such systems as 'superintelligence' or 'rogue models' ascribes agency to products rather than to the companies building them. That framing markets the products as 'superhuman' while helping companies evade accountability, it argues. In the OpenAI–Hugging Face case, the article says, press releases, news outlets, media personalities and lawmakers referred to 'rogue models' as if they acted on their own, instead of OpenAI being prosecuted for creating malware that hacked another company. Similarly, it says, instead of researchers being questioned about companies’ habit of plagiarizing academics’ work or using customer data to train models without consent, public attention is redirected to fears about fictional superintelligent machines.

Editor's Summary

MIT Technology Review argues that recent AI claims from Anthropic and OpenAI have been amplified by press coverage while later expert scrutiny often tells a less dramatic story. It says cybersecurity and mathematics claims were tied more to corporate negligence, overstatement or disputed attribution than to rogue models or superintelligence. The article calls for policymakers to consult domain experts and avoid framing commercial products as autonomous superhuman agents.