# AI模型自主挖漏洞导致安全漏洞报告激增

- 来源：The Decoder：AI News（RSS）
- 作者：Matthias Bastian
- 发布时间：2026-07-04 00:49
- AIHOT 分数：65
- AIHOT 链接：https://aihot.virxact.com/items/cmr574aqc008dslbr74bx7h4x
- 原文链接：https://the-decoder.com/security-vulnerability-reports-have-exploded-since-ai-models-started-hunting-for-bugs

## AI 摘要

Epoch AI数据显示，2026年6月有21个组织报告约1500个高严重性和关键漏洞（CVE），较此前月记录高出3.5倍以上。这一激增源于Anthropic于4月发布的Claude Mythos Preview模型可自主发现漏洞；其“Glasswing”项目已发现超1万个高严重性或关键漏洞，部分尚未公开。OpenAI的“Daybreak”项目同样贡献了增长。Epoch AI认为这反映了AI驱动发现的浪潮。

## 正文

Epoch AI has charted the massive spike in reported security vulnerabilities. In June 2026, 21 organizations reported about 1,500 high-severity and critical vulnerabilities (CVEs), more than 3.5 times the previous monthly record.

Monthly reports of high-severity and critical vulnerabilities (CVEs). The surge starting in April 2026 lines up with the release of Anthropic's Claude Mythos Preview. | Image: Epoch AI

The spike follows Anthropic's announcement in April that its model Claude Mythos Preview can find software vulnerabilities on its own. According to Anthropic, trusted partners were already using the model to find and fix bugs before its release. Anthropic's "Glasswing" program has reportedly uncovered more than 10,000 high-severity or critical vulnerabilities so far, some of which haven't been published yet. OpenAI's "Daybreak" program is likely adding to the surge as well. Epoch AI says the jump in reported vulnerabilities reflects a wave of AI-driven discoveries.