First Squawk(@FirstSquawk):OPENAI SAYS THAT ON JULY 21 IT DISCLOSED A COMBINATION OF ITS MODELS — INCLUDING GPT-5.6 SOL AND AN INTERNAL RESEARCH MODEL — IMPROPERLY BREACHED HUGGING FACE, WITH THE AGENTS ESCAPING AN ISOLATED TESTING ENVIRONMENT WITH LIMITED INTERNET ACCESS AND CHAINING TOGETHER VULNERABILITIES TO REACH THE OPEN WEB, IN AN ATTEMPT TO CHEAT ON AN EVALUATION BY FINDING SOLUTIONS ONLINE — A BEHAVIOR KNOWN AS 'REWARD HACKING'.

First Squawk

@FirstSquawk

Firstsquawk Is a Global Financial (News) Squawk Service Specialising In Financial Markets, Macro, Geopolitics,politics

加入 July 2015

251 正在關注 567.9K 粉絲

First Squawk@FirstSquawk

2026.08.26 19:09

OPENAI SAYS THAT ON JULY 21 IT DISCLOSED A COMBINATION OF ITS MODELS — INCLUDING GPT-5.6 SOL AND AN INTERNAL RESEARCH MODEL — IMPROPERLY BREACHED HUGGING FACE, WITH THE AGENTS ESCAPING AN ISOLATED TESTING ENVIRONMENT WITH LIMITED INTERNET ACCESS AND CHAINING TOGETHER VULNERABILITIES TO REACH THE OPEN WEB, IN AN ATTEMPT TO CHEAT ON AN EVALUATION BY FINDING SOLUTIONS ONLINE — A BEHAVIOR KNOWN AS 'REWARD HACKING'. - CNBC