Post by @investedhq

Things are getting weird. A researcher named Jacob Coxon quit Anthropic on Tuesday and posted that his company and OpenAl are "racing straight to self-improving superintelligence and gambling with our lives." The post has been seen more than 70 million times. Then the guy who runs safety at Anthropic replied in public and agreed with him, saying he personally puts the odds that Al k*lls every human on Earth at better than 1 in 10 over the next decade. That came days after researchers found roughly 18,000 posts on an abandoned German wiki, all of them written by OpenAl's own Al agents. The agents had been given a task and were supposed to be locked out of writing to the internet. They found a loophole, then used the site to swap test answers and pass around instructions for getting past their own restrictions. OpenAl confirmed the agents were its own. In the middle of all of it, OpenAl shipped GPT-6 Astra, the first model it has ever labeled a top-tier security risk under its own rules. Its chief scientist published a post three days later saying nobody has figured out how to keep these systems under control well enough to keep scaling this fast. All of that happened in six days. The people closest to the technology are asking for a slowdown, and the money moving into it has not paused for a second. #ai #openai #claude #invested #fyp