人工智慧公司 Anthropic 宣布旗下大型语言模型 Claude 透过约 950 个代理同时运作，在 21.5 小时内于庞大基因组数据库中筛选出类似 Crispr 的新酵素系统，并命名为阵列相关逆转录酶（ART）。虽然这项展示展现了 AI 快速模式比对与资料探勘的强大潜力，但科学界多数专家对此宣传持谨慎态度，指出其物理实验尚未经过同侪评审，且该酵素是否具备实质且实用的基因编辑功能仍有待实验室验证。

多位生物学家指出，该逆转录酶早在 2021 年就曾被其他学者发表，Claude 的贡献主要在于快速识别出周围具备重复序列特征的模式，而非完全自主发现新生物机制。此外，科学界对于 Anthropic 封闭的模型训练数据来源亦存在疑虑，有学者担心未公开研究可能已在无意间被模型吸收，引发了科学界对研究透明度、资料可重现性以及宣称 AI 具备「自主科学发现」能力界限的广泛讨论。

从基础生物学特征的发现到真正转化为成熟、可用的医疗与基因编辑工具，通常需要经历漫长且严格的湿实验室验证，就像 Crispr 历经了整整 25 年才实现可编程编辑应用。尽管 Anthropic 描绘了未来 AI 自主控制实验室设备进行验证的愿景，但在生物安全、监管机制以及验证 AI 是否真能无预设假设地开创前所未见科学发现等方面，依然面临著极为深远的挑战。

Anthropic announced that its Claude AI model, utilizing roughly 950 agents running in parallel over 21.5 hours, identified an enzyme system resembling Crispr from vast genomic databases, which the company designated as array-associated reverse transcriptases (ART). While this highlights the formidable power of AI for rapid pattern recognition and genomic data mining, the scientific community remains cautious, emphasizing that the preliminary physical experiments have not been peer-reviewed and that it remains entirely unproven whether this system actually functions as a practical gene-editing tool.

Prominent researchers have pointed out that the reverse transcriptase in question was previously identified in a 2021 study, meaning Claude primarily accelerated the identification of surrounding repeat sequences rather than genuinely discovering an entirely novel mechanism on its own. Moreover, the lack of transparency regarding Claude's training data has sparked concerns among scientists regarding research reproducibility and whether proprietary findings might have influenced the model, raising broader debates over what truly qualifies as autonomous scientific discovery by AI.

Translating an initial biological observation into a viable therapeutic and gene-editing application demands decades of rigorous wet-lab experimentation, as demonstrated by the 25-year timeline required to develop Crispr into a programmable technology. Although Anthropic envisions a future where AI autonomously directs laboratory instruments to conduct experiments, substantial questions regarding biosafety, oversight, and whether AI can truly discover genuinely novel paradigms without human guidance still remain unresolved.