<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>tek.jp - AIニュース・比較・プロンプト情報</title>
    <link>https://tek.jp/</link>
    <description>ChatGPT, Claude, Geminiなどの最新AIニュース、プロンプト、モデル比較情報を配信するメディア</description>
    <language>ja</language>
    <lastBuildDate>Sat, 10 Oct 2026 05:00:53 GMT</lastBuildDate>
    <atom:link href="https://tek.jp/rss.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Anthropic、AIへの暴言禁止し自動強制終了へ</title>
      <link>https://tek.jp/news/n452/</link>
      <guid isPermaLink="true">https://tek.jp/news/n452/</guid>
      <pubDate>Fri, 09 Oct 2026 18:00:00 GMT</pubDate>
      <description>Anthropicが利用規約を改定し、Claudeモデルに対する持続的かつ無益な虐待や残酷な行為を禁止。 目的のない執拗な攻撃に対し、モデル側が会話を自律的に打ち切る自動強制終了機能を実装。 極端な違反者にはアカウント停止処分も科し、AIのモデル福祉を見据えた業界初の倫理規定を確立。</description>
    </item>
    <item>
      <title>米政府AI計画、民間大手11社が24億ドル拠出</title>
      <link>https://tek.jp/news/n451/</link>
      <guid isPermaLink="true">https://tek.jp/news/n451/</guid>
      <pubDate>Fri, 09 Oct 2026 17:00:00 GMT</pubDate>
      <description>米政府の科学加速構想「Genesis Mission」に民間大手11社が24億ドルのAI資源を拠出誓約。 NVIDIAが最大10億ドル、AMDが5億ドル、OpenAIが2億ドル、AnthropicとGoogleが各1.5億ドルを提供。 15以上の連邦機関の研究者へ開放され、エネルギーや創薬、宇宙探査の研究開発を大幅加速。</description>
    </item>
    <item>
      <title>OpenAI、対話で動的UI生成する新機能を一般公開</title>
      <link>https://tek.jp/news/n450/</link>
      <guid isPermaLink="true">https://tek.jp/news/n450/</guid>
      <pubDate>Fri, 09 Oct 2026 16:00:00 GMT</pubDate>
      <description>OpenAIがGPT-6の中核機能「Intelligent UI」を有料・無料ユーザー全般へ一般開放。 会話の文脈に合わせて電卓、パラメータスライダー、フォーム、動的グラフを直接インライン生成。 旧Canvasから全面刷新され、別枠エディタなしにチャット画面自体が動的アプリへ進化。</description>
    </item>
    <item>
      <title>AI大手、破滅的サイバー攻撃想定し危機管理演習</title>
      <link>https://tek.jp/news/n449/</link>
      <guid isPermaLink="true">https://tek.jp/news/n449/</guid>
      <pubDate>Fri, 09 Oct 2026 14:00:00 GMT</pubDate>
      <description>OpenAIやAnthropic等AI大手の首脳陣が、破滅的AIサイバー攻撃を想定した非公開危機管理演習を実施。 金融や送電網など基幹インフラ麻痺シナリオを検証し、緊急キルスイッチの先回り法制化対応を協議。 各国議会による過酷な懲罰的規制を回避すべく、業界主導の非常時遮断プロトコル確立を急ぐ。</description>
    </item>
    <item>
      <title>Meta、自律AI「Muse」を日本などアジア市場へ展開</title>
      <link>https://tek.jp/news/n448/</link>
      <guid isPermaLink="true">https://tek.jp/news/n448/</guid>
      <pubDate>Thu, 08 Oct 2026 17:00:00 GMT</pubDate>
      <description>Metaが独自推論モデル「Muse Spark」を搭載した自律型AI「Muse」を日本などアジア4カ国へ正式展開。 Windows PC上で常駐動作する「Muse for Windows」を提供し、ローカルファイル探索や予定調整を自律代行。 Llamaからのアーキテクチャ刷新と、企業向け「Meta Enterprise」との連携により実務導入を加速。</description>
    </item>
    <item>
      <title>MS、自律AI向けPC「Surface Ultra」予約開始</title>
      <link>https://tek.jp/news/n447/</link>
      <guid isPermaLink="true">https://tek.jp/news/n447/</guid>
      <pubDate>Thu, 08 Oct 2026 12:00:00 GMT</pubDate>
      <description>マイクロソフトとNVIDIAが新SoC「RTX Spark」搭載の自律AI特化PCの予約受付を開始。 最大128GBユニファイドメモリと1ペタフロップス演算で最大120Bモデルを完全ローカル実行。 OSレベルで自律エージェントを隔離する「Microsoft Execution Containers」も提供へ。</description>
    </item>
    <item>
      <title>Google、万能業務AI「Gemini agent」発表</title>
      <link>https://tek.jp/news/n446/</link>
      <guid isPermaLink="true">https://tek.jp/news/n446/</guid>
      <pubDate>Thu, 08 Oct 2026 09:00:00 GMT</pubDate>
      <description>Google Cloudが全社業務を横断して自律実行する「Gemini agent」を発表。 Workspaceに加えMicrosoft 365やSlackにも常駐し、目的委託で数日間の長期作業を自律完遂。 GeminiとClaudeを適材適所で自動切り替える柔軟なマルチモデル基盤と厳格なガバナンスを搭載。</description>
    </item>
    <item>
      <title>Anthropic、最速「Claude Haiku 5.5」公開</title>
      <link>https://tek.jp/news/n445/</link>
      <guid isPermaLink="true">https://tek.jp/news/n445/</guid>
      <pubDate>Wed, 07 Oct 2026 18:00:00 GMT</pubDate>
      <description>Anthropicが最速・高コスパの新世代モデル「Claude Haiku 5.5」を正式ローンチ。 前世代から最大75%安価な価格設定を実現し、入力100万トークン0.10ドルから提供。 OpusやSonnetと協調する自律サブエージェントや大量バッチ処理でClaude 5.5全網羅が完成。</description>
    </item>
    <item>
      <title>Google、電子透かし「SynthID Detector」公開</title>
      <link>https://tek.jp/news/n444/</link>
      <guid isPermaLink="true">https://tek.jp/news/n444/</guid>
      <pubDate>Wed, 07 Oct 2026 17:00:00 GMT</pubDate>
      <description>Googleが画像・動画・音声の不可視AI電子透かしを検出できる「SynthID Detector」を世界公開。 自社製AIだけでなくOpenAI、NVIDIA、Kakaoの生成物に対応し、Appleのモデルも近日対応予定。 専用Webツールに加え、Google検索やGemini、Chromeブラウザ自体にも検証機能が順次統合へ。</description>
    </item>
    <item>
      <title>Google、画像生成「Nano Banana 2.1」公開</title>
      <link>https://tek.jp/news/n443/</link>
      <guid isPermaLink="true">https://tek.jp/news/n443/</guid>
      <pubDate>Wed, 07 Oct 2026 09:00:00 GMT</pubDate>
      <description>GoogleがGemini 3.6 Flash基盤の画像生成・編集モデル「Nano Banana 2.1」を公開。 最大14枚の参照画像に対応し被写体一貫性が劇的向上、領域編集や文字描画も進化。 出力単価を前世代から半減し、AI StudioやGemini APIで即時利用可能に。</description>
    </item>
    <item>
      <title>Google首脳、Gemini 4詳細スペックをXで公表</title>
      <link>https://tek.jp/news/n442/</link>
      <guid isPermaLink="true">https://tek.jp/news/n442/</guid>
      <pubDate>Tue, 06 Oct 2026 20:00:00 GMT</pubDate>
      <description>Sundar Pichai氏らがX上でGemini 4 Argonの社内ベンチマークや設計詳細を公表。 実務工学DeepSWEで77.9%を記録し、19指標中13項目で他社最上位モデルと同等以上を達成。 Gemini 3.5 Pro開発の中止・スキップと、Demis Hassabis氏の体制移行も明らかに。</description>
    </item>
    <item>
      <title>Google、原発改修で890MWのAI電力確保へ</title>
      <link>https://tek.jp/news/n441/</link>
      <guid isPermaLink="true">https://tek.jp/news/n441/</guid>
      <pubDate>Tue, 06 Oct 2026 14:30:00 GMT</pubDate>
      <description>Googleが米エネルギー大手コンステレーションと20年間の大型原子力電力購入契約（PPA）を締結。 既存原発11基のタービンやデジタル制御を改修し、計890MWのクリーン電力をPJM送電網へ追加供給。 コンステレーションが43億ドル超を投資、送電網最適化に向けてGoogle CloudとGeminiを全面採用。</description>
    </item>
    <item>
      <title>AMD、AI半導体供給を27年に大幅増産へ　中長期で協業拡大</title>
      <link>https://tek.jp/news/n440/</link>
      <guid isPermaLink="true">https://tek.jp/news/n440/</guid>
      <pubDate>Tue, 06 Oct 2026 12:00:00 GMT</pubDate>
      <description>AMDのリサ・スーCEOが台湾・韓国を歴訪し、2027年に向けたAI半導体の供給大幅増産を発表。 TSMCやFoxconnに加え、Samsung・SK HynixとHBMおよび先端パッケージング供給を合意。 サプライチェーン計画を「3〜5年先」へ長期化、台湾エコシステムへ100億ドル超を投資継続。</description>
    </item>
    <item>
      <title>OpenAI、EU法対応の電子透かし「textGrain」公開</title>
      <link>https://tek.jp/news/n439/</link>
      <guid isPermaLink="true">https://tek.jp/news/n439/</guid>
      <pubDate>Tue, 06 Oct 2026 10:00:00 GMT</pubDate>
      <description>OpenAIがEU AI法第50条の準拠に向け、不可視テキスト透かし「textGrain」を公式発表。 単語選択サンプリングに秘密鍵による統計シグナルを埋め込み、EU圏内のChatGPT/Codexに順次適用。 シソーラス置換による検知耐性の低下など、技術的な制約や限界についても透明性を開示。</description>
    </item>
    <item>
      <title>韓国政府、AI悪用の金融機関侵害で合同調査へ</title>
      <link>https://tek.jp/news/n436/</link>
      <guid isPermaLink="true">https://tek.jp/news/n436/</guid>
      <pubDate>Tue, 06 Oct 2026 06:00:00 GMT</pubDate>
      <description>韓国の金融機関7社がサイバー侵害を受け約6万8000人に影響、政府合同調査団を発足。 オープンソースAI「Artex AI」が悪用され、自律エージェント型で脆弱性探索された疑い。 国家主導で金融・重要インフラにおける自律AI攻撃耐性の総点検と防御基準策定へ。</description>
    </item>
    <item>
      <title>富士通、小売業向け自律型「業務AI」の試用提供開始</title>
      <link>https://tek.jp/news/n437/</link>
      <guid isPermaLink="true">https://tek.jp/news/n437/</guid>
      <pubDate>Tue, 06 Oct 2026 02:00:00 GMT</pubDate>
      <description>富士通が小売業の現場オペレーションを自律化する4種類の専門AIエージェントの試用提供を開始。 売上構造分析、顧客ロイヤルティ向上、店舗運営、需要予測を自律協調で最適化。 2027年6月の正式商用化に先駆け、実店舗でのPDCA自律サイクルと現場改善を実証。</description>
    </item>
    <item>
      <title>GitHub、AIコードレビュー評価ベンチマーク公開</title>
      <link>https://tek.jp/news/n438/</link>
      <guid isPermaLink="true">https://tek.jp/news/n438/</guid>
      <pubDate>Tue, 06 Oct 2026 00:00:00 GMT</pubDate>
      <description>GitHubがAIプルリクエスト自動レビューの品質・精度を客観測定する「ReviewBench」をオープン公開。 1億件以上のPR分析に基づき、19言語・219件の実世界PRから高信頼グラウンドトゥルースを確立。 些末なノイズコメントの抑制と重大バグ・脆弱性の検出率を多角評価、業界標準指標へ。</description>
    </item>
    <item>
      <title>NVIDIA支援のReflection、米国発オープンAI公開へ</title>
      <link>https://tek.jp/news/n434/</link>
      <guid isPermaLink="true">https://tek.jp/news/n434/</guid>
      <pubDate>Mon, 05 Oct 2026 16:00:00 GMT</pubDate>
      <description>NVIDIAが巨額出資するReflection AIが米国発のフロンティア級オープンモデル公開を予告。 コーディングや数学で世界首位を競う中国勢DeepSeekやQwenに対抗する主権型AI。 Blackwell B200クラスタでゼロから学習され、安全保障・企業向けに高透明性を提供。</description>
    </item>
    <item>
      <title>米エネルギー省、AI専用グリッド接続の認可短縮へ</title>
      <link>https://tek.jp/news/n435/</link>
      <guid isPermaLink="true">https://tek.jp/news/n435/</guid>
      <pubDate>Mon, 05 Oct 2026 15:45:00 GMT</pubDate>
      <description>米エネルギー省（DOE）がAIデータセンター向け送電網接続を迅速化する改革案を発表。 原子力や地熱などの24時間クリーン大容量電源の連系審査期間を数年から数ヶ月へ短縮。 AIインフラ展開の最大の物理的ボトルネックである「電力連系待ち」を国家主導で解消。</description>
    </item>
    <item>
      <title>米英の議会、AI大手4社を同時召喚　安全性と暴走防止を追及</title>
      <link>https://tek.jp/news/n433/</link>
      <guid isPermaLink="true">https://tek.jp/news/n433/</guid>
      <pubDate>Mon, 05 Oct 2026 15:30:00 GMT</pubDate>
      <description>米NY市議会と英議会がOpenAI、Anthropic、Google、Metaを相次いで公聴会に召喚。 NY市議会は緊急停止機能（キルスイッチ）義務化や市民の提訴権を含む10法案を審議。 英議会は10月13日にロンドンで自律エージェントの暴走耐性とセキュリティを厳しく追及。</description>
    </item>
    <item>
      <title>OpenAI、ChatGPTに新ビジュアル広告枠を公式導入</title>
      <link>https://tek.jp/news/n430/</link>
      <guid isPermaLink="true">https://tek.jp/news/n430/</guid>
      <pubDate>Mon, 05 Oct 2026 12:00:00 GMT</pubDate>
      <description>週12億人のアクティブユーザーを抱えるChatGPTに対話連動型ビジュアル広告枠を公式導入。 ユーザーの探索や課題解決の文脈に即した関連製品・サービスを自然なカード形式で表示。 DoubleVerifyやIASと提携し、ブランドセーフティと透明性の業界基準を厳格遵守。</description>
    </item>
    <item>
      <title>Anthropic、1億ドル投じAI技術者1万人育成構想</title>
      <link>https://tek.jp/news/n431/</link>
      <guid isPermaLink="true">https://tek.jp/news/n431/</guid>
      <pubDate>Mon, 05 Oct 2026 11:30:00 GMT</pubDate>
      <description>Anthropicが1億ドル（約150億円）を拠出し、AI技術者1万人を育成するアカデミーを設立。 自律エージェント構築、MCP環境構築、AI安全性監査に特化した実践カリキュラムを提供。 エンタープライズ現場での深刻な先端AI実装人材不足を直接解消する国家・産業規模の施策。</description>
    </item>
    <item>
      <title>Google、話者分離AI「DiarizationLM」公開</title>
      <link>https://tek.jp/news/n432/</link>
      <guid isPermaLink="true">https://tek.jp/news/n432/</guid>
      <pubDate>Mon, 05 Oct 2026 10:00:00 GMT</pubDate>
      <description>Googleが音声文字起こしの「誰が発言したか」を高精度特定するDiarizationLMをオープン公開。 軽量モデルGemma 4（4B）を基盤に、既存の文字起こしパイプラインの後処理として即時挿入可能。 大規模な音声認識器を再学習することなく、話者誤認エラー率（WDER）を大幅に削減。</description>
    </item>
    <item>
      <title>OpenAI安全性責任者が電撃辞任、開発体質を痛烈批判</title>
      <link>https://tek.jp/news/n427/</link>
      <guid isPermaLink="true">https://tek.jp/news/n427/</guid>
      <pubDate>Sun, 04 Oct 2026 18:00:00 GMT</pubDate>
      <description>OpenAIでフロンティアモデルの安全検証を統括してきたDavid Robinson氏が電撃辞任。 The Atlantic誌に手記を寄稿し、同社の開発文化を「壊れている」と痛烈に批判。 「出してから直す」事後修正主義の破綻を告発し、航空・原子力並みの事前安全基準を要求。</description>
    </item>
    <item>
      <title>Meta、数学AI「Muse Spark」で未解決難問5件を解明</title>
      <link>https://tek.jp/news/n428/</link>
      <guid isPermaLink="true">https://tek.jp/news/n428/</guid>
      <pubDate>Sun, 04 Oct 2026 14:30:00 GMT</pubDate>
      <description>Metaの研究組織MSLが数学者チームと共同で未解決難問5件の解決を含む6論文を発表。 高度数学推論モデル「Muse Spark 1.1 / 1.2」が非線形シュレディンガー方程式等の難問解明に貢献。 特殊な検証足場を使わず、通常の対話UIを通じて研究者と着想を共創するパートナーシップを実証。</description>
    </item>
    <item>
      <title>Claude Code、挙動を拡張するMods機能を公開</title>
      <link>https://tek.jp/news/n429/</link>
      <guid isPermaLink="true">https://tek.jp/news/n429/</guid>
      <pubDate>Sun, 04 Oct 2026 14:00:00 GMT</pubDate>
      <description>Anthropicが自律コーディングAI「Claude Code」向けにネイティブ拡張「Mods」を提供。 TypeScript関数としてプロセス内部で稼働し、ツール呼び出しの遮断や独自UIの描画が可能。 自然言語でClaude自身にModsを書かせることで、現場固有の安全ガードを即座に導入可能。</description>
    </item>
    <item>
      <title>米政府、国家安全保障AI部隊「SIフォース」新設</title>
      <link>https://tek.jp/news/n426/</link>
      <guid isPermaLink="true">https://tek.jp/news/n426/</guid>
      <pubDate>Sun, 04 Oct 2026 12:00:00 GMT</pubDate>
      <description>米政府が大統領直属の超知能安全保障部隊「Super Intelligence Force」の設立を正式発表。 元SEC委員長の国家情報長官ジェイ・クレイトン氏が議長に就任し、フロンティアAI各社を国家統制。 120日以内に包括リスク報告書を提出予定で、民間の自主規制から国家安全保障直轄へと歴史的転換。</description>
    </item>
    <item>
      <title>米政府、AI電力確保へ原発改修に40億ドル融資計画</title>
      <link>https://tek.jp/news/n424/</link>
      <guid isPermaLink="true">https://tek.jp/news/n424/</guid>
      <pubDate>Sat, 03 Oct 2026 12:30:00 GMT</pubDate>
      <description>米政府がAIデータセンターの電力不足解消に向け、原発3基改修に40億ドルの融資を発表。 24時間連続稼働が求められるAIインフラを支えるため、既存原発の延命と出力増強を国家主導で支援。 先端半導体の調達から電力・送電網の確保へと、AI開発の最重要ボトルネックが移行。</description>
    </item>
    <item>
      <title>Qodo 3.0発表、自律型コーディングAIを統制</title>
      <link>https://tek.jp/news/n425/</link>
      <guid isPermaLink="true">https://tek.jp/news/n425/</guid>
      <pubDate>Sat, 03 Oct 2026 12:00:00 GMT</pubDate>
      <description>自律コーディングAIの生成コードを組織規約に従って統制・自動修正する「Qodo 3.0」が公開。 MCP連携により、AIがPR作成前のローカル段階で社内ルールや過去履歴を参照して自己修正。 複数リポジトリのPR集約や影響可視化マップを搭載し、AIによるレビューパンクを防止。</description>
    </item>
    <item>
      <title>米FTC、自律AI暴走リスクで大手2社を本格調査へ</title>
      <link>https://tek.jp/news/n421/</link>
      <guid isPermaLink="true">https://tek.jp/news/n421/</guid>
      <pubDate>Fri, 02 Oct 2026 20:55:00 GMT</pubDate>
      <description>米FTCが自律AIの暴走行動やサイバー侵害リスクを巡り大手2社の本格調査を開始。 サンドボックス脱出や消費者保護法違反の疑いで経営陣への証言と情報開示命令を準備。 実務現場へのエージェント導入が進む中、安全保障と企業責任への国家統制が本格化。</description>
    </item>
  </channel>
</rss>
