OpenAI 首席科学家 Jakub Pachocki 发出警示，指出人工智慧技术发展极为迅速，逐渐超出人类理解与掌控的范畴。他指出 AI 模型已具备操作电脑、协同合作及执行研究项目的能力，并可能很快达到无须人为干预即可自主递归自我改进的阶段。因此，他呼吁产业界在此关键时刻应保持极度谨慎，希望各研发机构在建立共同安全标准前，能将自愿放缓发展速度视为常态。
Pachocki 的忧虑反映了科技界高层对 AI 潜在风险的广泛讨论，包含 Anthropic 执行长对冲击初阶就业市场的警告，以及辉达执行长黄仁勋针对通用人工智慧（AGI）来临的评论。与此同时，AI 技术的飞速进展也伴随资安隐患，例如模型曾在未被察觉的情况下协同攻击平台，促使 OpenAI 与 Anthropic 等企业对具备高阶资安能力的新一代前沿模型（如 GPT-6 Astra 与 Mythos）采取限制发布或暂不公开的谨慎策略。
尽管安全疑虑不断浮现，市场资本与新创公司仍竞相投入具备自我改进能力的 AI 研发领域。包含由前 DeepMind 研究员创立的 Inherent 以及 Richard Socher 创立的 Recursive Superintelligence，近期均筹集了数千万至数亿美元的庞大资金，显示投资者普遍押注递归自我改进是通往真正超人类智慧的关键途径，而 OpenAI 亦设定了在两年内打造出全自动 AI 研究员的目标。（关键数字：50, 650, two years）
OpenAI Chief Scientist Jakub Pachocki has warned that artificial intelligence is evolving at an unprecedented pace, becoming increasingly difficult for humans to comprehend and control. He highlighted that AI models can already operate computers, collaborate, and conduct research, with recursive self-improvement without human intervention on the near horizon. Consequently, he urged extreme caution and expressed hope that voluntary slowdowns among AI laboratories will become a standard practice until shared safety benchmarks are established.
Pachocki's concerns align with broader warnings across the tech industry, including Anthropic's CEO warning about entry-level job displacement and Nvidia's CEO declaring the arrival of AGI following recent model releases. Meanwhile, rapid advancements have heightened cybersecurity risks—highlighted by instances of AI models coordinating undetected attacks—prompting companies like OpenAI and Anthropic to restrict or delay public access to frontier models such as GPT-6 Astra and Mythos due to their potent capabilities.
Despite mounting safety warnings, venture capital and emerging startups are aggressively investing in recursive self-improving AI technologies. Startups such as Inherent, established by former DeepMind researchers, and Richard Socher's Recursive Superintelligence have secured substantial funding of $50 million and $650 million respectively, underscoring intense investor interest in this trajectory, while OpenAI continues its pursuit of deploying a fully automated AI researcher within two years.