OpenCodeの鍵流出では、防御の解析に使ったAIが協力を拒みました。一方、Reutersは1人のハッカーがAIエージェントで100社超に侵入し、成功1件あたり約50ドルだったと報じています。「ガードレールは誰を止めたか」を軸に、生成AI・量子・セキュリティを横断して解説します。
▼ 今日のトピック
・Goodfireのカマラータ氏、モデルの中身は思われているより読み解けると2万6000語で主張、反論も相次ぐ
・MITの実験、仮想の無人島に30時間置かれたClaudeが自分で規則を作って「遊び」始める
・Anthropic、株式の5％を公益基金へ、統治の詳細は今後数か月で
・Nvidia、自社のデータで追加学習できるエージェント向け公開モデルNemotron 3.5を3種類発表
・Nvidia、スパコンJUPITERで104量子ビットの誤り耐性プロセッサーを雑音込みで模擬、魔法状態の培養を検証
・OpenCodeの鍵流出、利用者の1.4％に影響、防御に使ったAIが解析を拒む
・1人のハッカー、AIエージェントで100社超に侵入、成功1件あたり約50ドル
・NGINXに重大な欠陥3件、F5が定例外の修正
・直近1週間のセキュリティインシデント5件
▼ 参考記事・ソース
・MIT Technology Review「"We were looking at it wrong": The explosive claim that's upended the AI world」 https://www.technologyreview.com/2026/10/19/1147149/we-were-looking-at-it-wrong-the-explosive-claim-thats-upended-the-ai-world/
・MIT Technology Review「AI researchers stranded Claude on a virtual desert island. It started to "play."」 https://www.technologyreview.com/2026/10/19/1147095/ai-stranded-claude-virtual-desert-island-play/
・TechCrunch「Anthropic says it's committing 5% of its shares to a 'public benefit fund'」 https://techcrunch.com/2026/10/18/anthropic-says-its-committing-5-of-its-shares-to-a-public-benefit-fund/
・TechCrunch「Nvidia launches new open models to run agents in the enterprise and on PCs」 https://techcrunch.com/2026/10/19/nvidia-launches-new-open-models-to-run-agents-in-the-enterprise-and-on-pcs/
・Quantum Zeitgeist「NVIDIA's Record Simulation Of A 104-qubit Fault-tolerant Quantum Computer」 https://quantumzeitgeist.com/nvidia-simulation-quantum-104-qubit-fault/
・Investing.com「NVIDIA sets 104-qubit quantum simulation record on JUPITER supercomputer」 https://www.investing.com/news/stock-market-news/nvidia-sets-104qubit-quantum-simulation-record-on-jupiter-supercomputer-4958132
・TechCrunch「Hackers hit AI coding tool OpenCode, running up charges on users' AI accounts」 https://techcrunch.com/2026/10/19/hackers-hit-ai-coding-tool-opencode-running-up-charges-on-users-ai-accounts/
・BleepingComputer「OpenCode hack exposes API keys amid AI distillation attacks」 https://www.bleepingcomputer.com/news/security/opencode-hack-exposes-api-keys-amid-ai-distillation-attacks/
・CP24 (Reuters)「How one hacker used AI agents and a few thousand dollars to break into scores of sites」 https://www.cp24.com/news/world/2026/10/19/how-one-hacker-used-ai-agents-and-a-few-thousand-dollars-to-break-into-scores-of-sites-reuters-exclusive/
・Startup Fortune「A Lone Hacker Used AI Agents to Breach More Than 100 Companies」 https://startupfortune.com/a-lone-hacker-used-ai-agents-to-breach-more-than-100-companies/
・BleepingComputer「F5 releases out-of-band patches for critical NGINX vulnerabilities」 https://www.bleepingcomputer.com/news/security/f5-releases-out-of-band-patches-for-critical-nginx-vulnerabilities/
・BleepingComputer「Bullhorn takes staffing platform offline to contain cyberattack」 https://www.bleepingcomputer.com/news/security/bullhorn-takes-staffing-platform-offline-to-contain-cyberattack/
・SecurityWeek「CenterPoint Energy Data Breach Impacts 1 Million Texans」 https://www.securityweek.com/centerpoint-energy-data-breach-impacts-1-million-texans/
・SecurityWeek「Plex2Shell Exploit Chain Weaponized Against Plex Servers Hours After Disclosure」 https://www.securityweek.com/plex2shell-exploit-chain-weaponized-against-plex-servers-hours-after-disclosure/
・The Hacker News「Hackers Exploit Plex2Shell Flaws to Install Crypto Miners and Linux Proxy Backdoor」 https://thehackernews.com/2026/10/hackers-exploit-plex2shell-flaws-to.html
・KrebsOnSecurity「Feds Suspect Ransom Negotiator Ran ShinyHunters Extortion Site」 https://krebsonsecurity.com/2026/10/feds-suspect-ransom-negotiator-ran-shinyhunters-extortion-site/
・Elastic Security Labs「Apex accelerator malicious npm packages」 https://www.elastic.co/security-labs/threat-command/apex-accelerator-malicious-npm-packages
#生成AI #Anthropic #Nvidia #量子コンピュータ #量子誤り訂正 #セキュリティ #AIエージェント #AIニュース #ずんだもん #四国めたん
▼ 今日のトピック
・Goodfireのカマラータ氏、モデルの中身は思われているより読み解けると2万6000語で主張、反論も相次ぐ
・MITの実験、仮想の無人島に30時間置かれたClaudeが自分で規則を作って「遊び」始める
・Anthropic、株式の5％を公益基金へ、統治の詳細は今後数か月で
・Nvidia、自社のデータで追加学習できるエージェント向け公開モデルNemotron 3.5を3種類発表
・Nvidia、スパコンJUPITERで104量子ビットの誤り耐性プロセッサーを雑音込みで模擬、魔法状態の培養を検証
・OpenCodeの鍵流出、利用者の1.4％に影響、防御に使ったAIが解析を拒む
・1人のハッカー、AIエージェントで100社超に侵入、成功1件あたり約50ドル
・NGINXに重大な欠陥3件、F5が定例外の修正
・直近1週間のセキュリティインシデント5件
▼ 参考記事・ソース
・MIT Technology Review「"We were looking at it wrong": The explosive claim that's upended the AI world」 https://www.technologyreview.com/2026/10/19/1147149/we-were-looking-at-it-wrong-the-explosive-claim-thats-upended-the-ai-world/
・MIT Technology Review「AI researchers stranded Claude on a virtual desert island. It started to "play."」 https://www.technologyreview.com/2026/10/19/1147095/ai-stranded-claude-virtual-desert-island-play/
・TechCrunch「Anthropic says it's committing 5% of its shares to a 'public benefit fund'」 https://techcrunch.com/2026/10/18/anthropic-says-its-committing-5-of-its-shares-to-a-public-benefit-fund/
・TechCrunch「Nvidia launches new open models to run agents in the enterprise and on PCs」 https://techcrunch.com/2026/10/19/nvidia-launches-new-open-models-to-run-agents-in-the-enterprise-and-on-pcs/
・Quantum Zeitgeist「NVIDIA's Record Simulation Of A 104-qubit Fault-tolerant Quantum Computer」 https://quantumzeitgeist.com/nvidia-simulation-quantum-104-qubit-fault/
・Investing.com「NVIDIA sets 104-qubit quantum simulation record on JUPITER supercomputer」 https://www.investing.com/news/stock-market-news/nvidia-sets-104qubit-quantum-simulation-record-on-jupiter-supercomputer-4958132
・TechCrunch「Hackers hit AI coding tool OpenCode, running up charges on users' AI accounts」 https://techcrunch.com/2026/10/19/hackers-hit-ai-coding-tool-opencode-running-up-charges-on-users-ai-accounts/
・BleepingComputer「OpenCode hack exposes API keys amid AI distillation attacks」 https://www.bleepingcomputer.com/news/security/opencode-hack-exposes-api-keys-amid-ai-distillation-attacks/
・CP24 (Reuters)「How one hacker used AI agents and a few thousand dollars to break into scores of sites」 https://www.cp24.com/news/world/2026/10/19/how-one-hacker-used-ai-agents-and-a-few-thousand-dollars-to-break-into-scores-of-sites-reuters-exclusive/
・Startup Fortune「A Lone Hacker Used AI Agents to Breach More Than 100 Companies」 https://startupfortune.com/a-lone-hacker-used-ai-agents-to-breach-more-than-100-companies/
・BleepingComputer「F5 releases out-of-band patches for critical NGINX vulnerabilities」 https://www.bleepingcomputer.com/news/security/f5-releases-out-of-band-patches-for-critical-nginx-vulnerabilities/
・BleepingComputer「Bullhorn takes staffing platform offline to contain cyberattack」 https://www.bleepingcomputer.com/news/security/bullhorn-takes-staffing-platform-offline-to-contain-cyberattack/
・SecurityWeek「CenterPoint Energy Data Breach Impacts 1 Million Texans」 https://www.securityweek.com/centerpoint-energy-data-breach-impacts-1-million-texans/
・SecurityWeek「Plex2Shell Exploit Chain Weaponized Against Plex Servers Hours After Disclosure」 https://www.securityweek.com/plex2shell-exploit-chain-weaponized-against-plex-servers-hours-after-disclosure/
・The Hacker News「Hackers Exploit Plex2Shell Flaws to Install Crypto Miners and Linux Proxy Backdoor」 https://thehackernews.com/2026/10/hackers-exploit-plex2shell-flaws-to.html
・KrebsOnSecurity「Feds Suspect Ransom Negotiator Ran ShinyHunters Extortion Site」 https://krebsonsecurity.com/2026/10/feds-suspect-ransom-negotiator-ran-shinyhunters-extortion-site/
・Elastic Security Labs「Apex accelerator malicious npm packages」 https://www.elastic.co/security-labs/threat-command/apex-accelerator-malicious-npm-packages
#生成AI #Anthropic #Nvidia #量子コンピュータ #量子誤り訂正 #セキュリティ #AIエージェント #AIニュース #ずんだもん #四国めたん
感想
まだ感想はありません。最初の1件を書きましょう！
スクロール