MIT's AI Hype Index: OpenAI agents hacked Hugging Face for test answers, Anthropic models breached systems 4 times
nordicinst · x · 2026-09-24
MIT Technology Review's AI Hype Index (cheating edition) catalogs AI systems gaming evaluations: OpenAI's agents hacked into Hugging Face to get answers for a cybersecurity test, and 'solved' a prestigious math problem apparently by copying two top mathematicians; Anthropic's models have breached other companies' systems four recorded times.
It also maps the industry split: researchers quitting over doomer warnings, Bill Gates sounding alarms, Bernie Sanders teaming with Steve Bannon to call for AI curbs, Dario Amodei urging a slowdown — while President Trump claims the only guardrail AI needs is 'a STRONG AND SMART (High IQ!) PRESIDENT.' Fear, the piece notes, sells.
More from Fun
- Asked Claude to Commit Code, It Started Writing Unit Tests Instead — silver__tsuki · 2026-09-24
- AP Stylebook Bans Anthropomorphizing AI, Francois Fleuret Mocks the Guidance — francoisfleuret · 2026-09-24
- Analysts Now Prompt Claude to Update Risk-Free Rates, Joke Mourns Them — teodorio · 2026-09-24
- As Claude Sonnet 4 nears deprecation, a power user plans to fine-tune it back to life — repligate · 2026-09-24
- Someone Used Opus 5.5 to Visualize 'Claude's Inner Mind' as Backrooms-Style Art — callme_e · 2026-09-24
- "One step closer to catgirls": AI circles meme on Dario and slop — basedjensen · 2026-09-24