Recent hacking and math announcements drew fanfare, but independent experts found negligence, overstatement, and plagiarism.
Over the past few months, Anthropic and OpenAI made dramatic claims about hacking, math, and superintelligence. Press coverage often repeated company framings, but independent experts later described the hacking incidents as basic security failures and the math results as less novel than advertised. Mathematicians accused OpenAI of misconduct and plagiarism.
Anthropic claimed its Claude Mythos model outperformed most security experts at finding vulnerabilities, and OpenAI faced a hacking incident involving Hugging Face. Anthropic and Meta then disclosed similar incidents. Anthropic later claimed a math breakthrough, and OpenAI said its chatbot Astra solved problems open for a decade. New York University professor Tristan Buckmaster disputed OpenAI’s attribution.
Builders and operators should separate vendor marketing from verified capability, especially when security and math benchmarks arrive with anthropomorphic language. The hacking incidents point to neglected security practices rather than rogue agents, and the math disputes show that benchmark claims need independent replication before teams adopt them.
Watch for expert responses, security incident writeups, and corrections from the companies involved. Those documents will show whether the summer’s claims hold up and how quickly enterprises should adjust procurement, model evaluation, and incident response plans.
What matters
- Anthropic and OpenAI promoted hacking and math claims that experts later disputed.
- Builders should treat vendor security and benchmark claims as marketing until independent review arrives.
- Watch for expert rebuttals and detailed incident reports from security researchers and mathematicians.
Why it matters
Watch for expert rebuttals and detailed incident reports from security researchers and mathematicians.
This GenAI News article was prepared in original wording using reporting and materials published by MIT Technology Review AI. Source reference: https://www.technologyreview.com/2026/09/22/1144867/dont-be-fooled-summer-ai-hype/.
Drafted by the GenAI News review pipeline.