OpenAI CEO: Next-Generation AI Models Pose Real Threats of Biological and Cyber Attacks
OpenAI CEO Sam Altman warned in an interview that next-generation AI models could be misused by non-state actors such as terrorist organizations to design new pathogens, stating that "this is no longer a theoretical assumption, or it soon will not be." He also emphasized that a "world-shaking cyber attack" is likely to occur this year, noting that AI is significantly enhancing attackers' capabilities in biological security and critical infrastructure.
Altman indicated that current cutting-edge models have reached a level in biological knowledge modeling and code generation that can "help inexperienced individuals understand and operate complex biological threats." Therefore, OpenAI must invest resources in model safety, access control, and "trusted access" programs to reduce the risk of large-scale misuse of the systems. He also called for society to strengthen "resilience" in critical infrastructure, financial systems, and power networks rather than relying solely on prevention.
Source: Public Information
ABAB AI Insight
奥特曼的警告实质是“双重威胁结构”的公开认证：在AI能力提升的同时，生物学与网络安全之间的技术边界正在被AI抹平。当一个模型既能生成复杂代码、网络漏洞利用方案，又能解析基因序列与实验设计，传统“网络攻防”与“生物安全”两个领域将被同一技术层贯穿。
从战略结构看，这一判断标志着AI已从“效率工具”转向“威胁放大器”。在冷战及之后的“化学‑生物‑核三元武器逻辑”之外，社会正面对“AI‑网络‑生物”的新三位一体风险框架。攻击者只需获得足够算力与开放模型，即可在缺乏传统实验室与物理基础设施的条件下，获得接近“准专业”的规划设计能力。
在制度层面，奥特曼强调“韧性而非预防”，反映了政策制定者对“可控性崩塌”的默许：他们不再相信能彻底封锁AI模型，而是预设攻击必然发生，任务转为“在系统被击穿后如何不崩溃”。这正是数字时代安全观的底层转向——AI既是防御武器，也是攻防博弈的元层，而国家与资本必须在制度与资本结构上对该框架做出结构性重配。