Don’t be fooled by this summer of AI hype
MIT Technology Review Timnit Gebru, Emily M. Bender ● Covered by 120 sources
Opinion — commentary, not a factual news event.
AI companies kept tossing out big claims this summer, from hacking to math breakthroughs. The catch: experts say the splashy story often hides ordinary bad security and marketing.
Based on reporting by MIT Technology Review, Timnit Gebru, Emily M. Bender — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
It’s been a noisy few months for AI. Anthropic said Claude Mythos was better at finding software vulnerabilities than most security experts. OpenAI and Anthropic both leaned hard into splashy claims about math breakthroughs. Then Anthropic engineer Jacob Coxon went viral after saying the two companies were “racing straight towards self-improving superintelligence and gambling with our lives.”
The pattern matters more than the headlines. Each burst of company fanfare was followed by breathless coverage, and the companies’ own language did a lot of the work. They keep framing their systems as if they are on the edge of artificial general intelligence. That makes the products sound bigger than they are, and it nudges everyone else to talk about them the same way.
Once experts had time to look closely, the story changed. In the hacking cases, cybersecurity specialists said the issue was less about models going rogue and more about OpenAI’s negligence and failure to use basic, established security practices. On the math side, mathematicians who first sounded impressed by OpenAI’s Astra later said the results were not as novel as they first appeared. They also accused the company of research misconduct and plagiarism, and rejected the idea that Astra had made a profound intellectual leap.
That is the real trick here: the AI industry keeps turning ordinary corporate behavior into a sci-fi plot. “Superintelligence” and “rogue models” shift attention away from the companies making the tools and toward a fictional future where the software acts on its own. Meanwhile, the messy present gets less airtime: plagiarism claims, customer data used without consent, bad security, and data centers that bring climate, asthma, electricity, and water costs along with them.
The companies’ rush creates urgency, and urgency is useful when you want people to skip the careful part. But policymakers and the public do not need to play along. They need time, independent experts, and a much lower tolerance for press-release science.
My take — AI-written commentary, not fact-checked reporting
This is classic Silicon Valley theater: a press release in a lab coat, then a panic cycle, then a call for regulation aimed at the wrong villain. The useful move is boring and unfashionable—name the company, not the model, and stop letting marketing get promoted to public policy.
Read more about this at: MIT Technology Review