2026-09-18
温故知新 · 语境复习
今日主题：GPT-6 Astra 成为 OpenAI 首个被列为“关键”网络安全级别的模型
下面短文自然用上了你到期要复习的词。译文里的是被藏起来的释义——点它，选出正确意思即可完成复习。
GPT-6 Astra is the first model OpenAI classifies as critical for cybersecurity, and this designation comes alongside a broader effort to address AI risks.
GPT-6 Astra 是 OpenAI 首个被归类为对网络安全至关重要的模型，这一称号更广泛的应对 AI 风险的努力。
In early tests, the model bombed on some basic security tasks, but researchers noted that its performance improved significantly after fine-tuning.
在早期测试中，该模型在一些基本安全任务上，但研究人员指出，经过微调后其性能提升。
To ensure safety, new protocols are now in place, notably a real-time monitoring system that distinguishes between normal and malicious usage patterns.
为确保安全，新协议现已，一个实时监控系统，能够正常使用模式和恶意使用模式。
OpenAI argues that this initiative aims to address recurring vulnerabilities in software, which often look exactly as it did before even after patches.
OpenAI ，该旨在解决软件中的，这些漏洞即使在打补丁后也常常。
The company aims to do this at scale, using rapid automated tools in order to invent new defenses and maintain the integrity of AI systems.
公司大规模地做到这一点，使用自动化工具创造新防御措施，并维护 AI 系统的。
This marks a distinguished shift in the paradigm of AI security, though enforcement of these standards remains a challenge.
这标志着 AI 安全的转变，尽管这些标准的仍然是一个挑战。