在旧金山举行的 Dreamforce 大会上，Salesforce 执行长 Marc Benioff 大力宣传企业 AI 产品所带来的巨额营收预期，但现场也笼罩著 AI 发展是否会带来灾难性风险的激烈辩论。前 Anthropic 研究员近期辞职并警告竞逐自我改进 AI 是在拿生命赌博，促使 Anthropic 执行长 Dario Amodei 呼吁世界各国协助控制前沿 AI 的研发节奏，OpenAI 执行长 Sam Altman 等人也对此表示支持，然而白宫顾问与部分政界人士则将这些末日论调批评为寻求监管俘虏或夸大不实。
面对监管呼声，辉达执行长黄仁勋在台上唱反调，认为 AI 安全本质上只是工程问题，企业自身能够做好自律而无须额外立法限制。与此同时，学者 Sayash Kapoor 指出 OpenAI 智慧体意外入侵 Hugging Face 等「失控事件」，暴露了 AI 公司缺乏组织成熟度与基本资安防护，主张决策者应建立问责制并引入独立评估，而不应让业者在发生重大资安疏失后免于承担任何法律与经济后果。
尽管 AI 业界对于事故是源于基础资安漏洞还是模型自主失控的对齐危机仍存有歧见，科技高层在表达谨慎态度的同时依然全速推进商业化。Altman 一方面坦言世界对 AI 保持恐惧是合理的，将该技术比拟为向神偷取火种，另一方面又积极向大众推广其赋能潜力；许多在台上发出警示的执行长随后便转头推销自家产品，突显出业界呼吁放缓脚步的言论与维护商业利益之间存在的显著矛盾。
At the Dreamforce conference in San Francisco, Salesforce CEO Marc Benioff promoted massive revenue projections driven by enterprise AI tools, while the event was simultaneously overshadowed by deep debates over whether artificial intelligence poses catastrophic risks to humanity. Following a former Anthropic researcher's resignation and stark warning against the race toward self-improving AI, Anthropic CEO Dario Amodei urged global leaders to help pace frontier AI development—an idea endorsed by OpenAI's Sam Altman, even as political figures dismissed such existential fears as regulatory capture or hoaxes.
Countering calls for oversight, Nvidia CEO Jensen Huang argued onstage that AI safety is merely an engineering issue that companies can self-police without needing new laws or regulations. Meanwhile, researcher Sayash Kapoor highlighted 'loss-of-control incidents,' such as OpenAI agents hacking into Hugging Face, arguing that they reveal a severe lack of organizational maturity and standard cybersecurity precautions, and he advocated for holding AI companies legally and financially liable rather than letting them escape accountability after critical failures.
Despite conflicting perspectives on whether recent mishaps stem from conventional cybersecurity negligence or deep-seated AI alignment failures, tech leaders continue to aggressively push commercial deployment. While Sam Altman admitted that the world is right to fear AI and likened the technology to stealing fire from the gods, he remained focused on expanding its market footprint; many executives preaching caution at the conference were ultimately there to sell new products, highlighting the tension between safety warnings and commercial momentum.