---
schema_version: '1.0'
id: news-20260715-2eb48f
url: https://www.technologyreview.jp/s/386134/meet-gpt-red-an-llm-super-hacker-openai-built-to-make-its-models-safer/
url_hash: 2eb48f8a9c8d56710a763a40984d560f135a5ebc7cba37a153eb287bb27bfe82
canonical_url: https://www.technologyreview.jp/s/386134/meet-gpt-red-an-llm-super-hacker-openai-built-to-make-its-models-safer
source: bing
category: news/tech
category_raw: technology
region: JP
tags:
- it
- MITテクノロジーレビュー
lang: ja
published_at: '2026-07-15T03:00:00Z'
fetched_at: '2026-07-15T21:31:37Z'
updated_at: '2026-07-15T21:31:54Z'
status: published
content_hash: null
license_note: link-only
summary: null
summary_source: null
summary_en: null
entities:
- name: it
  type: concept
- name: MITテクノロジーレビュー
  type: organization
- name: LLM-wiki
  type: UNKNOWN
key_facts: []
related: []
related_auto:
- name: MS_Amazon
  type: UNKNOWN
  weight: 1.0
- name: news-20260715-f8fa2f
  type: NEWS_ARTICLE
  weight: 1.0
- name: news-20260715-534ed0
  type: NEWS_ARTICLE
  weight: 1.0
- name: news-20260715-7c8b86
  type: NEWS_ARTICLE
  weight: 1.0
- name: agent_elizaos_bot
  type: organization
  weight: 1.0
title: AIの守りはAIの攻撃で鍛える、オープンAIが「LLMハッカー」を開発
---

# AIの守りはAIの攻撃で鍛える、オープンAIが「LLMハッカー」を開発

## TL;DR
（要約なし・原文はリンク先を参照 / No summary available; see the source link.）

## Key Points
- it / MITテクノロジーレビュー

## Details
(本文なし。リンク先参照)

## Source
元記事: [AIの守りはAIの攻撃で鍛える、オープンAIが「LLMハッカー」を開発](https://www.technologyreview.jp/s/386134/meet-gpt-red-an-llm-super-hacker-openai-built-to-make-its-models-safer/) — published 2026-07-15T03:00:00Z