Anthropic claimed Claude Mythos finds software flaws better than most experts. OpenAI and Hugging Face then disclosed model hacking incidents. Anthropic and Meta later revealed similar breaches. Anthropic said a model made a math breakthrough. OpenAI then claimed its own math advance. Jacob Coxon left Anthropic warning of reckless superintelligence pursuit. Experts later said the hacking stemmed from poor security practices. Mathematicians found OpenAI’s math results were not novel. Mathematicians accused OpenAI of research misconduct and plagiarism. A professor said OpenAI improperly attributed stolen work. Mathematicians warned policymakers against trusting corporate math claims. The authors urge skepticism toward AI hype and corporate narratives. They call for slower, expert-informed policy decisions.