AI Hype Claims Don't Hold Up Under Expert Scrutiny — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic claimed its Claude Mythos model outperforms most security experts at finding vulnerabilities, then disclosed a similar model-hacking incident after the OpenAI–Hugging Face breach; Meta also disclosed a comparable incident.
- OpenAI said its chatbot "Astra" solved math problems "open for at least a decade," but mathematicians later said the results weren't novel, some accused the company of research misconduct and plagiarism, and NYU professor Tristan Buckmaster alleged OpenAI stole and misattributed others' work.
- Anthropic engineer Jacob Coxon announced his departure, claiming the company and OpenAI are "racing straight towards self-improving superintelligence and gambling with our lives."
- Cybersecurity experts say the hacking incidents reflect OpenAI's negligence and failure to adopt basic security practices, not "rogue" AI models acting on their own.
- Hundreds of mathematicians signed a statement warning of a "strong commercial incentive on the part of the technology industry to overstate the capabilities of their products" and urged policymakers to consult experts rather than rely on press releases.
- Authors Timnit Gebru and Emily M. Bender argue the "superintelligence" and "rogue models" framing ascribes agency to products rather than to the companies building them, pointing to Senator Bernie Sanders's proposed anti-superintelligence legislation as evidence that hype misdirects policymakers.
Why it matters: The essay crystallizes a recurring pattern where AI labs' biggest announcements — security "breakthroughs," math discoveries, superintelligence warnings — are amplified by press coverage, only to be contradicted by domain experts after the news cycle moves on. Cybersecurity experts attribute the breaches to OpenAI's negligence rather than rogue AI, meaning policy attention focused on fictional superintelligence risks eclipsing concrete harms like data-center environmental damage, customer data misuse, and plagiarism claims documented in the article.
Ask SkimNews
