• Home
  • Services
  • About
  • Contact
  • Blog
  • 知財活動のROICへの貢献
  • 生成AIを活用した知財戦略の策定方法
  • 生成AIとの「壁打ち」で、新たな発明を創出する方法

​
​よろず知財コンサルティングのブログ

オムロン、知財AIエージェントを内製し特許関連工数を50%削減

3/9/2026

0 Comments

 
2026年9月2日付の日経クロステックの記事「オムロンがBedrock活用で知財AIエージェント内製、特許関連工数を50%減」によれば、オムロンは、AWSの生成AI基盤「Amazon Bedrock」などを活用し、先行技術調査と発明説明書作成を支援する知財AIエージェントを自社開発し、2026年1月頃から研究開発部門で利用を始め、これらの業務に要する工数を、少なく見積もっても50%削減したということです。
オムロンでは、発明考案のプロセスを、①発明の創出、②先行技術調査、③発明説明書の作成、の3段階に整理しています。エンジニアが本来注力したいのは発明の創出ですが、専門的な特許文献の確認や説明書作成に多くの時間を取られていました。
そこで、特許庁から取得した過去の出願データをAmazon S3に格納し、Amazon OpenSearch ServiceとAmazon Bedrockを組み合わせました。AIエージェントは、入力されたキーワードや仕様書を基に関連文献を検索・要約し、類似技術の整理、競合分析、発明説明書の作成まで支援します。
商用データベースやSaaSの利用も検討しましたが、データ量の制約や独自の業務プロセスへの対応を考え、内製を選択しました。また、出願前の機密情報を安全に扱うため、AWS上に共通AI基盤「RDinX」を構築し、セキュリティー設定やAIエージェントの挙動、プロンプトを一元的に管理しています。
今回の事例は、生成AIの活用が単なる文書要約から、発明発掘、調査、分析、説明書作成をつなぐ業務プロセスの再設計へ進んでいることを示しています。AIに判断を任せ切るのではなく
、人が結果を確認しながら、エンジニアを定型作業から解放し、発明創出に集中させる取り組みとして注目されます。
この記事の内容を生成AIに深掘り調査させましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
オムロンがBedrock活用で知財AIエージェント内製、特許関連工数を50%減
2026.09.02
https://active.nikkeibp.co.jp/atcl/act/19/00798/080300002/
 
 
OMRON Develops In-House IP AI Agent, Cutting Patent-Related Workload by 50%
According to a Nikkei XTECH article published on September 2, 2026, titled “OMRON Develops In-House IP AI Agent Using Bedrock, Cutting Patent-Related Workload by 50%,” OMRON has developed its own intellectual property AI agent using Amazon Web Services’ generative AI platform, Amazon Bedrock, and other AWS services. The agent supports prior-art searches and the preparation of invention disclosure documents. OMRON began using it within its research and development divisions around January 2026 and has conservatively estimated that it has reduced the time required for these tasks by at least 50%.
OMRON divides the invention development process into three stages: (1) generating an invention, (2) conducting a prior-art search, and (3) preparing an invention disclosure document. Although engineers would ideally devote most of their attention to creating inventions, they had been spending considerable time reviewing highly technical patent documents and preparing disclosure documents.
To address this issue, OMRON stored historical patent application data obtained from the Japan Patent Office in Amazon S3 and combined Amazon OpenSearch Service with Amazon Bedrock. Based on keywords or technical specifications entered by users, the AI agent searches for and summarizes relevant documents. It also assists with organizing similar technologies, analyzing competitors, and preparing invention disclosure documents.
OMRON considered using commercial databases and software-as-a-service solutions. However, it chose to develop the system in-house because of restrictions on data volume and the need to accommodate its proprietary business processes. To securely handle confidential information before patent applications are filed, the company also built a shared AI platform called “RDinX” on AWS. The platform centrally manages security settings, AI agent behavior, and prompts.
This case illustrates how the use of generative AI is evolving beyond simple document summarization toward the redesign of entire workflows connecting invention discovery, prior-art research, analysis, and the preparation of disclosure documents. Rather than leaving decisions entirely to AI, OMRON has adopted a human-in-the-loop approach in which people review the AI-generated results. The initiative is particularly noteworthy because it frees engineers from routine tasks and enables them to concentrate more fully on creating inventions.
I asked generative AI to conduct an in-depth investigation of the matters discussed in this article. Please refer to the results below. Please note, however, that the research and analysis generated by AI are based solely on publicly available information. They may not necessarily reflect the actual circumstances and may contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

タイ、英国、韓国、中国、日本のソブリンAI

3/9/2026

0 Comments

 
世界各国の人工知能(AI)戦略が民間主導の技術開発から、国家が主導する社会実装へと移行しています。生成AIに、タイ、英国、韓国、中国、日本の5カ国を対象に調査させましたので、ご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
Sovereign AI in Thailand, the United Kingdom, South Korea, China, and Japan
Artificial intelligence (AI) strategies around the world are shifting from private-sector-led technology development toward government-led implementation across society. I asked generative AI to examine the initiatives of five countries—Thailand, the United Kingdom, South Korea, China, and Japan—and have shared the results for your reference.
Please note that the research and analysis conducted by generative AI are based solely on publicly available information. They may not necessarily reflect the actual situation and may contain inaccuracies.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

「Claude Fable 5.1」と「Claude Mythos 5.1」発表

2/9/2026

0 Comments

 
2026年9月1日、Anthropicは最新モデル「Claude Fable 5.1」と「Claude Mythos 5.1」を発表しました。両モデルは、コーディング、AIエージェント、科学研究などの高度なタスクを対象としたモデルです。大きなポイントは、Fable 5.1とMythos 5.1が完全に別のモデルではなく、同じ基盤モデルを異なるセーフガードで提供していることです。
AIモデルの性能を測定する独立評価プラットフォーム「Artificial Analysis」が提供する、AIモデルの総合的な知能を評価するための複合指標(ベンチマーク)「Artificial Analysis Intelligence Index」では、66で最高評価です。(Claude Opus 5 (max) 63, GPT-5.6 Sol (max) 61)
「コーディングと知識労働のための世界最先端のモデルであり、その研究能力は、AIモデルが科学の進歩にどのように貢献していくかを示す初期段階の展望を示しています。」ということですので、楽しみです。
Claude Fable 5.1 と Claude Mythos 5.1について生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
Claude: Fable 5.1 and Mythos 5.1
https://www.anthropic.com/claude-fable-and-mythos-5-1
 
 
Anthropic Announces “Claude Fable 5.1” and “Claude Mythos 5.1”
On September 1, 2026, Anthropic announced its latest models, Claude Fable 5.1 and Claude Mythos 5.1. Both models are designed for advanced tasks such as coding, AI agent operations, and scientific research. One of the most notable aspects of the announcement is that Fable 5.1 and Mythos 5.1 are not entirely separate models; rather, they are based on the same underlying foundation model but are offered with different safeguards.
On the Artificial Analysis Intelligence Index—a composite benchmark developed by Artificial Analysis, an independent AI model evaluation platform, to assess the overall intelligence of AI models—the models received the highest score of 66. By comparison, Claude Opus 5 (max) scored 63, while GPT-5.6 Sol (max) scored 61.
Anthropic describes the model as “the world’s most advanced model for coding and knowledge work,” adding that its research capabilities offer an early glimpse of how AI models may contribute to scientific progress. This is certainly something to look forward to.
I asked generative AI to conduct an in-depth analysis of Claude Fable 5.1 and Claude Mythos 5.1, and I invite you to read the results. Please note that this research and analysis were generated by AI based solely on publicly available information. They may not necessarily reflect the actual situation and may contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

Gensparkが描く「仕事を終わらせるAI」

2/9/2026

0 Comments

 
2026年8月30日付のForbes JAPANの記事「AIの勝負はモデルでは終わらない、Genspark創業者が描く『次のAIサービス』の条件」を読みました。
この記事が示しているのは、生成AIの競争軸が、ChatGPT、Gemini、Claudeなどの「どのモデルが最も賢いか」という競争から、その知能を誰でも無理なく使い、実際の仕事の完了までつなげられるかという競争へ移り始めていることです。AIモデルの能力は大幅に向上しましたが、利用者が自分の立場、目的、過去の経緯を何度も説明し、複数のツールを行き来しなければならない仕事の流れは、まだ十分には変わっていないとGensparkの創業者たちは見ています。
Gensparkの考え方を象徴するのが、OpenAIやAnthropicなどが高性能な「エンジン」を作る企業であるのに対し、Gensparkは利用者を目的地まで運ぶ「車」を作る企業だ、という役割分担です。利用者がモデルの性能を比較して使い分けるのではなく、情報収集、判断、文書作成、共有、次の作業までを一続きにし、「どのモデルを使うか」ではなく「何を終わらせるか」を中心にサービスを設計しようとしています。
その中核となる構想が「SecondBrain」です。メール、会議、チャット、文書、アプリ、プロジェクト履歴などを継続的な文脈としてAIに持たせ、AIをその都度質問に答えるだけの道具から、利用者の仕事を長期的に理解する存在へ変えようとしています。Genspark 6.0では、このSecondBrainを基盤として、スーパーエージェントや各種作成ツール、複数のAIエージェントを組み合わせ、「記憶」「判断」「生成」「協働」を一つの流れとして結び付けています。
さらに、Gensparkが目指しているのは、AIをテキスト入力欄の中だけに閉じ込めないことです。会議の会話、移動中のメモ、対面での打ち合わせなど、これまで十分にデジタル化されていなかった情報も仕事の文脈として取り込み、その後の整理、連絡、資料作成、意思決定へつなげようとしています。最終的な目標は、「Googleを使える人ならGensparkも使える」というほど、専門的なプロンプト技術を必要としないAIです。
知財業務に引き寄せて考えると、今後のAIツールは、特許検索や要約、クレーム解釈の精度だけでは評価できなくなると考えられます。案件の背景、過去の検索式、審査経過、競合企業の動向、事業部門との議論などを継続的に理解し、調査から分析、報告書作成、意思決定支援までをつなげられるかが重要になります。一方で、AIが多くの業務文脈を保持するほど、機密情報の管理、アクセス権限、情報の正確性、判断過程の記録、人間による最終確認の重要性も高まります。これは、便利なAIツールを導入するという話を超え、企業の業務プロセスそのものを再設計する問題です。
「AIの勝負はモデルでは終わらない」という記事タイトルは、生成AI市場が「賢い回答を返すAI」から、「文脈を理解し、仕事を実際に前へ進めるAI」へ移行していることを端的に表しています。
このGensparkが描く「次のAIサービス」の条件と、AI産業、企業の業務変革、知財業務への影響について生成AIに深掘りさせましたので、ご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
AIの勝負はモデルでは終わらない、Genspark創業者が描く「次のAIサービス」の条件
8/30(日)
https://news.yahoo.co.jp/articles/bfde0325786588fe3ba4f650d0505ec046cad97c
 
 
Genspark’s Vision of “AI That Gets the Job Done”
I read the Forbes JAPAN article published on August 30, 2026, titled “The AI Race Does Not End with Models: Genspark’s Founders on What the ‘Next Generation of AI Services’ Requires.”
The article suggests that the focus of competition in generative AI is beginning to shift away from the question of which model—ChatGPT, Gemini, Claude, or another—is the most intelligent. Instead, the competition is increasingly about who can make that intelligence effortlessly accessible to everyone and use it to bring actual work through to completion. AI models have become dramatically more capable, but Genspark’s founders believe that workflows have not yet changed sufficiently: users still have to repeatedly explain their position, objectives, and previous circumstances, while moving back and forth between multiple tools.
Genspark’s philosophy is symbolized by a clear division of roles. While companies such as OpenAI and Anthropic build high-performance “engines,” Genspark sees itself as building the “car” that takes users to their destination. Rather than requiring users to compare models and decide which one to use, Genspark aims to connect information gathering, judgment, document creation, sharing, and subsequent tasks into a single continuous workflow. Its services are therefore designed around “what the user wants to get done,” rather than “which model should be used.”
At the heart of this vision is a concept called “SecondBrain.” Genspark aims to give AI ongoing access to contextual information from emails, meetings, chats, documents, applications, and project histories. This is intended to transform AI from a tool that merely answers individual questions into an entity that understands the user’s work over the long term. In Genspark 6.0, SecondBrain serves as the foundation for combining its Super Agent, various creation tools, and multiple AI agents, integrating “memory,” “judgment,” “generation,” and “collaboration” into a single workflow.
Genspark also aims to prevent AI from being confined to a text input box. It seeks to incorporate information that has not previously been sufficiently digitized—such as conversations during meetings, notes made while traveling, and face-to-face discussions—into the context of a user’s work. That information can then be connected to subsequent organization, communication, document preparation, and decision-making. The ultimate goal is to create an AI system that requires no specialized prompting skills—one that is so easy to use that “anyone who can use Google can use Genspark.”
Viewed from the perspective of intellectual property work, future AI tools will no longer be evaluated solely on the accuracy of patent searches, summaries, or claim interpretation. What will matter is whether they can continuously understand the background of a matter, previous search queries, prosecution histories, competitors’ activities, and discussions with business divisions, and then connect research, analysis, report preparation, and decision-making support into one coherent process.
At the same time, the more business context an AI system retains, the more important it becomes to manage confidential information, control access permissions, verify the accuracy of information, maintain records of the decision-making process, and ensure final human review. This goes beyond merely introducing a convenient AI tool. It raises the broader question of how companies should redesign their business processes themselves.
The article’s title, “The AI Race Does Not End with Models,” succinctly captures the generative AI market’s transition from “AI that provides intelligent answers” to “AI that understands context and actually moves work forward.”
I asked generative AI to conduct an in-depth examination of the conditions that Genspark believes the “next generation of AI services” must satisfy, as well as the implications for the AI industry, corporate business transformation, and intellectual property work. Please refer to the resulting analysis. Please note, however, that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect actual circumstances, and may contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

韓国政府「独自AI基盤モデル」2次評価の詳細点数を発表

2/9/2026

0 Comments

 
2026年8月27日、韓国政府(科学技術情報通信省)は、国家代表の人工知能(AI)を選ぶ「独自AI基盤モデル」プロジェクトの2次段階評価の詳細点数を発表し、各評価項目について参加企業すべての得点を明らかにしました。
SKテレコムが総合1位を獲得した一方で、国際的な性能指標で首位だったMotif Technologiesが脱落したことを受けたものです。
生成AIに本件を深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
韓国政府、「独自AI基盤モデル」2次評価の全点数公開…総合1位はSKテレコム
8/31(月)
https://news.yahoo.co.jp/articles/5ab59f2485118c999dcef6460965197cc1cf58af
 
 
South Korean Government Releases Detailed Scores from the Second-Round Evaluation of Its “Sovereign AI Foundation Model” Project
On August 27, 2026, the South Korean government’s Ministry of Science and ICT announced the detailed results of the second-round evaluation for its “Sovereign AI Foundation Model” project, which is intended to select a nationally representative artificial intelligence model. The ministry disclosed the scores received by every participating company for each evaluation criterion.
The detailed scores were released following the outcome in which SK Telecom ranked first overall, while Motif Technologies was eliminated despite having achieved the highest score on international performance benchmarks.
I asked generative AI to conduct an in-depth analysis of this matter, and the results are presented below for your reference. Please note that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the actual circumstances, and may contain inaccurate information.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

防衛省の情報分析をSakana AIが支援

2/9/2026

0 Comments

 
2026年8月28日付のLedge.aiの記事「防衛省の情報分析をSakana AIがAI支援 公開情報で収集・分析・管理、AIエージェント技術を実証」を読みました。
Sakana AIは、防衛省と「総合分析業務に必要なAI機能の調査・実証」に関する契約を締結しました。契約額は約3億2,900万円で、AIエージェント技術などを活用し、情報収集の効率化、分析能力の向上、体系的な情報管理を検証します。
防衛省の総合分析業務では、政策や部隊運用の意思決定に必要な情報を、公開情報、電波情報、画像・地理情報などから収集・分析します。情報量が急増する一方、生成AIにはハルシネーションや情報管理上の課題があるため、単なる要約ではなく、情報源の確認や人間による検証を含む仕組みが重要になります。
今回の実証対象は、有料・無料を問わない「公開情報」です。したがって、秘密情報や電波情報をSakana AIが直接分析するものではなく、新聞、ウェブ、論文、統計などを活用するOSINTの高度化が中心とみられます。言語モデルだけでなく、エージェント型AI、推薦型AI、マルチモーダルAI、オンプレミス環境などの組み合わせも検討されます。
Sakana AIは2026年3月にも、防衛装備庁から指揮統制システム関連の研究を受注していますが、今回の案件は、安全保障政策全体を支える戦略的な情報分析が対象です。AIが分析官に代わるのではなく、膨大な公開情報の整理や関連性の発見を支援し、人間の意思決定を高度化できるかが注目されます。
この記事の内容を基に、生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
防衛省の情報分析をSakana AIがAI支援 公開情報で収集・分析・管理、AIエージェント技術を実証
https://ledge.ai/articles/sakana_ai_defense_information_analysis
 
 
Sakana AI Supports the Ministry of Defense’s Intelligence Analysis
I read an August 28, 2026 article published by Ledge.ai titled “Sakana AI to Support the Ministry of Defense’s Intelligence Analysis with AI—Demonstrating AI Agent Technology for the Collection, Analysis, and Management of Open-Source Information.”
Sakana AI has signed a contract with Japan’s Ministry of Defense for a project titled “Research and Demonstration of AI Capabilities Required for Comprehensive Analysis Operations.” The contract is valued at approximately ¥329 million. Through the use of AI agent technologies and other tools, the project will examine ways to make information collection more efficient, enhance analytical capabilities, and manage information systematically.
In its comprehensive analysis operations, the Ministry of Defense collects and analyzes the information needed for decision-making on policy and military operations from sources including publicly available information, signals intelligence, imagery, and geospatial information. While the volume of available information is growing rapidly, generative AI also presents challenges, including hallucinations and information-management risks. A system that goes beyond simple summarization—and incorporates source verification and human review—will therefore be essential.
The scope of this demonstration project is limited to “publicly available information,” regardless of whether access is free or paid. Sakana AI will therefore not directly analyze classified information or signals intelligence. Instead, the project appears to focus primarily on enhancing open-source intelligence, or OSINT, by using newspapers, websites, academic papers, statistical data, and other sources. In addition to language models, the project may examine combinations of agentic AI, recommendation systems, multimodal AI, and on-premises computing environments.
In March 2026, Sakana AI also received a research contract from Japan’s Acquisition, Technology & Logistics Agency concerning command-and-control systems. The latest project, however, focuses on strategic intelligence analysis supporting Japan’s broader national security policy. The key question is not whether AI will replace intelligence analysts, but whether it can help them organize vast amounts of publicly available information, identify previously unnoticed connections, and improve the quality of human decision-making.
Based on the contents of this article, I asked generative AI to conduct a more in-depth analysis. Please note that the resulting research and analysis are based solely on publicly available information. They may not necessarily reflect the actual circumstances and could contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

韓国・知識財産処「バイブコーディング業務革新」

1/9/2026

0 Comments

 
韓国の知識財産処(旧・特許庁)は、2026年8月27日に「バイブコーディング(Vibe Coding)ワークショップ」を開催し、生成AIを活用した業務革新の取り組みを本格化させました。専門的なプログラミング知識がなくても、自然語(日常の言葉)で開発意図を説明するだけで業務ツールを構築できる「バイブコーディング」により、行政業務の効率化とデジタル転換(DX)を加速させています。
この韓国の動きを生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
「言葉で説明すれば業務アプリ完成」…知識財産先、「バイブコーディング」でAI業務革新
https://m.news.nate.com/view/20260827n32612
 
 
South Korea’s Intellectual Property Authority Promotes “Business Innovation through Vibe Coding”
On August 27, 2026, South Korea’s intellectual property authority, formerly the patent office, held a “Vibe Coding Workshop,” marking the full-scale launch of its efforts to transform administrative operations through generative AI. Vibe coding enables users to build business tools simply by describing what they want to develop in natural, everyday language, without requiring specialized programming knowledge. Through this approach, the authority aims to improve the efficiency of administrative work and accelerate digital transformation (DX).
I asked generative AI to conduct a more in-depth investigation and analysis of these developments in South Korea. Please note, however, that the AI-generated research and analysis are based solely on publicly available information and may not necessarily reflect the actual circumstances. They may also contain inaccuracies. Please keep this in mind when reviewing the material.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

8月のAIニュース

1/9/2026

0 Comments

 
YouTube「【月刊AIニュース】8月のAIニュース46本をこの一本で全部解説」は、ちょっと長い(約1時間51分)ですが、8月のAI関連ニュースを良くまとめています。
この動画の内容を基に、生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
【月刊AIニュース】8月のAIニュース46本をこの一本で全部解説
https://www.youtube.com/watch?v=3lWH8rL3lIQ
 
 
AI News from August
The YouTube video, “[Monthly AI News] All 46 AI News Stories from August Explained in One Video,” is rather long—approximately one hour and 51 minutes—but provides an excellent overview of AI-related news from August.
Based on the content of this video, I asked generative AI to conduct a more in-depth investigation and analysis. Please note, however, that the results generated by AI are based solely on publicly available information and may not necessarily reflect the actual circumstances. They may also contain inaccuracies. Please keep this in mind when reviewing the material.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

PerplexityがNVIDIAと提携した Portable Computer

1/9/2026

0 Comments

 
2026年8月25日、Perplexityは、新しいローカルエージェントプラットフォーム「Portable Computer」をNVIDIA DGX Sparkに最適化することで、ユーザーのローカルハードウェア上においてパーソナルAIエージェントを実行できるようにすると発表しました。
本製品は、推論モデルや実行環境をユーザーのローカルハードウェアへ移行する「ローカルファースト」設計を採用し、データ主権の確保と運用コストの削減を同時に実現しています。動作にはNVIDIA DGX Sparkなどの高性能な計算基盤を必要とし、通常タスクは端末内で完結させつつ、高度な判断が必要な場合のみクラウドへ委譲するハイブリッド型の仕組みが特徴です。独自のPPLX 27Bモデルや決定論的なハーネスコードを組み合わせることで、小型モデルながら既存のオープンソース環境を凌駕する高い処理精度を実証しています。機密データを扱う企業やヘビーユーザーに対し、プライバシーと経済性を両立した新たな形態を提示する内容となっています。
この「Portable Computer」を生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
Perplexity、DGX Sparkに最適化されたローカルAIエージェントをリリース
https://ai.watch.impress.co.jp/docs/news/2136082.html
 
 
Perplexity’s “Portable Computer,” Announced in Partnership with NVIDIA
On August 25, 2026, Perplexity announced that it would enable users to run personal AI agents on their own local hardware by optimizing its new local agent platform, “Portable Computer,” for NVIDIA DGX Spark.
The platform adopts a “local-first” design that moves inference models and execution environments onto the user’s local hardware, simultaneously ensuring data sovereignty and reducing operating costs. It requires a high-performance computing platform such as NVIDIA DGX Spark and employs a hybrid architecture in which routine tasks are completed entirely on the local device, while only tasks requiring more advanced judgment are delegated to the cloud. By combining its proprietary PPLX 27B model with deterministic harness code, Perplexity has demonstrated high processing accuracy that surpasses existing open-source environments despite the model’s relatively compact size. The platform presents a new approach that combines privacy with cost efficiency for enterprises handling confidential data and for power users.
I asked generative AI to conduct an in-depth analysis of this “Portable Computer,” and I invite you to refer to the results. Please note, however, that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the actual circumstances, and may contain inaccuracies.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

知財訴訟の実際―当事者は何を考え、なぜ訴訟に至るのか

31/8/2026

0 Comments

 
2026年8月28日に行われた「【第4回サマリア知財フォーラム】知財訴訟の実際 -当事者は何を考え、訴訟に至るのか?-【講師】河部 康弘 氏: 河部法律事務所 所長/パテント・インテグレーション株式会社 顧問弁護士」のアーカイブ動画(約62分」を視聴しました。
パテント・インテグレーション株式会社の顧問弁護士でもある河部康弘氏が、豊富な訴訟経験を踏まえ、知財紛争における当事者の心理と意思決定を解説しています。河部氏が実質的に関与した、裁判所の事件番号が付された案件は約120件に上るとのことです。
日本では、勝訴者が実際に支払った弁護士費用を、そのまま敗訴者に負担させる制度にはなっていません。裁判には時間も費用もかかるため、権利者も被警告者も、本音では訴訟を避けながら、自らに有利な条件で解決したいと考えます。
権利者は「要求に応じなければ訴える」と圧力をかけ、被警告者は「訴えるならどうぞ」と強気に対応します。双方が相手の譲歩を待ち続け、ブレーキが効かなくなった結果、訴訟に突入する――これが講演でいう「チキンレース」です。
権利者が訴訟を提起するかどうかは、勝訴可能性だけでは決まりません。得られる損害賠償額、回収可能性、無効主張を受けるリスク、訴訟費用、社内負担、顧客やサプライチェーンへの影響、さらに経営者や発明者の感情まで含めて判断されます。
特に重要なのは、知財訴訟も事業活動の一環であるという点です。侵害を放置すれば、「この会社は権利行使をしない」と市場に受け取られ、他社の模倣を誘発するおそれがあります。一方で、相手の事業規模が小さく、自社への影響も限定的であれば、費用を考えて訴えない判断もあり得ます。
被警告者側も、侵害・無効の判断だけでなく、「相手は本当に訴えてくる会社か」を見ています。強力な無効資料を示したり、対象製品がほとんど売れていないことを開示したりして、権利者の訴訟意欲を低下させることも重要な交渉戦略になります。
講演で特に興味深かったのは、特許の実効的な権利範囲は、クレームの文言だけでは決まらないという指摘です。
過去の訴訟実績、訴訟予算の確保、訴訟費用保険への加入、知財訴訟に詳しい弁護士との連携などにより、「この会社は本当に訴えてくる」と相手に認識させれば、訴訟前の交渉力が高まります。強い明細書を作るだけでなく、必要な場合に権利行使できる体制と覚悟を示すことも、特許の牽制力を高める知財戦略といえます。
訴訟開始後は、裁判所の心証が示されることで侵害や無効の見通しが立ちやすくなり、中立的な裁判官が双方の事情を調整するため、和解も成立しやすくなります。
知財戦略は、特許を取得して終わりではありません。権利行使を実行できる組織、予算、社内意思決定、外部専門家との関係まで準備しておくことが、保有特許の価値と企業の交渉力を支える――非常に示唆に富む講演でした。
 
動画はこちらです。
https://www.youtube.com/watch?v=G9OetwbtBRE
 
 
The Reality of IP Litigation—What Do the Parties Think, and Why Do Disputes End Up in Court?
I watched the approximately 62-minute archived video of the seminar held on August 28, 2026, entitled “The 4th Summaria IP Forum: The Reality of IP Litigation—What Do the Parties Think, and Why Do Disputes End Up in Court?” The speaker was Yasuhiro Kawabe, Managing Partner of Kawabe Law Office and Legal Advisor to Patent Integration Co., Ltd.
Drawing on his extensive litigation experience, Mr. Kawabe, who also serves as legal advisor to Patent Integration Co., Ltd., explained the psychology and decision-making processes of parties involved in IP disputes. According to him, he has been substantially involved in approximately 120 cases that were formally assigned court case numbers.
In Japan, the losing party is not generally required to reimburse the prevailing party for the full amount of attorneys’ fees actually incurred. Because litigation is both time-consuming and costly, both rights holders and recipients of infringement warnings would, in truth, prefer to avoid going to court while still resolving the dispute on terms favorable to themselves.
The rights holder applies pressure by saying, “Comply with our demands, or we will sue,” while the accused party responds defiantly, “Go ahead and sue us.” Each side continues waiting for the other to back down, until the brakes cease to function and the dispute proceeds to litigation. This is what Mr. Kawabe described in the seminar as a “game of chicken.”
Whether a rights holder initiates litigation is not determined solely by the likelihood of prevailing. The decision also takes into account the potential amount of damages, the likelihood of actually recovering them, the risk of facing an invalidity challenge, litigation costs, the internal burden on the company, the impact on customers and supply chains, and even the emotions of management and the inventors involved.
One particularly important point is that IP litigation is itself part of business activity. If infringement is left unchallenged, the market may conclude that “this company does not enforce its rights,” potentially encouraging other companies to imitate its products or technology. On the other hand, when the other party’s business is small and the impact on the rights holder is limited, the rights holder may reasonably decide not to sue in view of the costs involved.
The recipient of an infringement warning also considers not only infringement and validity issues, but whether the rights holder is genuinely the kind of company that will bring a lawsuit. Presenting strong prior art that could invalidate the patent, or disclosing that the accused product has generated almost no sales, can reduce the rights holder’s incentive to litigate. This can therefore be an important negotiation strategy.
One of the most interesting observations in the seminar was that the effective scope of patent rights is not determined solely by the wording of the claims.
A company can strengthen its pre-litigation bargaining position by making the other party believe that it is genuinely prepared to sue. Such credibility may be established through a history of prior litigation, the allocation of an adequate litigation budget, the purchase of litigation expense insurance, and close cooperation with attorneys experienced in IP litigation. An effective patent strategy therefore involves not only drafting strong patent specifications, but also demonstrating that the company has both the organizational capability and the determination to enforce its rights when necessary. This can substantially enhance the deterrent effect of its patents.
Once litigation has commenced, the court may indicate its preliminary assessment, making it easier for the parties to evaluate the likely outcomes regarding infringement and validity. In addition, a neutral judge can take both parties’ circumstances into account and help bridge their differences, making settlement more likely.
An IP strategy does not end when a patent is granted. The value of a company’s patent portfolio and its negotiating power depend on advance preparation—including an organization capable of enforcing rights, an appropriate budget, internal decision-making procedures, and established relationships with external specialists.
It was an exceptionally insightful seminar.
 
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

自衛隊の通信網にNTT「IOWN」の中核技術導入

31/8/2026

0 Comments

 
防衛省は、全国の基地や部隊を結ぶ自衛隊の通信網に、NTTの次世代光通信基盤「IOWN(アイオン)」の中核技術であるオールフォトニクス・ネットワーク(APN)を導入する方針を固め、次期防衛力整備計画に明記し、2027年度予算案に関連費を計上、2028年度以降の本格運用を目指すということです。
この件について生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
自衛隊情報基盤にIOWN導入 AI活用へ大容量通信、次期計画に明記
2026/8/29
https://www.nikkei.com/article/DGXZQOUA276CL0X20C26A8000000/
 
 
Japan Self-Defense Forces to Introduce NTT’s Core IOWN Technology into Communications Network
Japan’s Ministry of Defense has reportedly decided to introduce the All-Photonics Network (APN)—a core technology of NTT’s next-generation optical communications infrastructure, IOWN—into the communications network connecting Japan Self-Defense Forces bases and units nationwide. The plan is expected to be incorporated into the next Defense Buildup Program, with related expenditures included in the fiscal 2027 budget proposal and full-scale operation targeted for fiscal 2028 or later.
I asked generative AI to conduct an in-depth investigation into this development. Please refer to the findings below. Please note, however, that the research and analysis generated by AI are based solely on publicly available information and may not necessarily reflect the actual circumstances. They may also contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

OpenAIが2026年末までに社内でAGIを実現する

30/8/2026

0 Comments

 
2026年8月26日TIME誌公開のインタビュー記事で、OpenAIのサム・アルトマンCEOは「2026年末までに社内でAGI(汎用人工知能)を実現するという予測を示しました。
この件について、生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
OpenAIは2026年末までにAGIと呼ばれるシステムを開発するだろうとサム・アルトマンCEOが回答
https://gigazine.net/news/20260827-openai-agi/
 
 
OpenAI to Achieve AGI Internally by the End of 2026
In an interview published by TIME on August 26, 2026, OpenAI CEO Sam Altman predicted that the company would achieve artificial general intelligence (AGI) internally by the end of 2026.
I asked generative AI to conduct an in-depth investigation into this topic. Please note that the resulting research and analysis are based solely on publicly available information, may not necessarily reflect the actual circumstances, and may contain inaccuracies.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

AIが機器を操作する共通仕様MHS

30/8/2026

0 Comments

 
Paragraph. 編集するにはここをクリック.​8月30日 AIが機器を操作する共通仕様MHS
 
Anthropicは2026年8月27日(現地時間)、AIエージェントが物理的な機器を安全に操作するための共通仕様「Model Hardware Standard」(MHS)の研究プレビューを、最初の科学研究機関および先進的な製造業者グループに公開すると発表しました。MHSにより、AIエージェントは顕微鏡、液体ハンドラー、ロボットアームなどの複数の実験機器や製造機器を並行して操作し、日常的な創薬実験から量子コンピュータ上のレーザー校正まで、複雑なタスクを実行できるようになります。
Anthropicが2024年に公開した「MCP (Model Context Protocol)」がAIエージェントとツールやデータを接続するためのプロトコルなのに対し、MHSは物理デバイスを扱うための仕様です。
この「Model Hardware Standard」(MHS)について生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
AIと機器接続の共通規格開発 米アンソロピック、時間短縮へ
8/28(金)
https://news.yahoo.co.jp/articles/04807df85975846361a9d4bc1662c0b617c30dd4
 
 
MHS: A Common Standard for AI to Operate Physical Equipment
On August 27, 2026, local time, Anthropic announced that it would release a research preview of the Model Hardware Standard (MHS)—a common specification designed to enable AI agents to operate physical equipment safely—to an initial group of scientific research institutions and advanced manufacturers.
MHS will enable AI agents to operate multiple pieces of laboratory and manufacturing equipment in parallel, including microscopes, liquid-handling systems, and robotic arms. This will allow them to perform complex tasks ranging from routine drug-discovery experiments to laser calibration on quantum computers.
Whereas Anthropic’s Model Context Protocol (MCP), released in 2024, is a protocol for connecting AI agents to tools and data, MHS is a specification for interacting with physical devices.
I asked generative AI to conduct an in-depth analysis of the Model Hardware Standard (MHS). Please see the results below. Please note that the research and analysis produced by generative AI are based solely on publicly available information and may not necessarily reflect the actual situation. They may also contain inaccuracies.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

Tencent「Hy4 preview」発表

29/8/2026

0 Comments

 
中国のテック大手Tencentは、2026年8月28日に新世代のフラッグシップ大規模言語モデル「Hy4 preview」(テンセント混元/Hunyuan Hy4)を発表し、オープンソース(オープンウェイト)として一般公開しました。
Hy4 previewはMixture-of-Experts、略してMoEという方式で、すべての計算部分を毎回動かすのではなく、入力に応じて必要な専門部分だけを動かします。総パラメータ数は7700億ですが、1トークンの処理で稼働するのは490億。
このHy4-previewについて生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。

​Tencent Announces “Hy4 Preview”
Chinese technology giant Tencent announced its next-generation flagship large language model, “Hy4 Preview”—also known as Tencent Hunyuan Hy4—on August 28, 2026, and made it publicly available as an open-source, or more precisely open-weight, model.
Hy4 Preview employs a Mixture-of-Experts (MoE) architecture. Rather than activating the entire model for every computation, it selectively activates only the expert components needed for each input. Although the model has 770 billion parameters in total, only 49 billion parameters are activated when processing each token.
I asked generative AI to conduct an in-depth analysis of Hy4 Preview, and the results are provided for your reference. Please note, however, that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the actual situation, and may contain inaccuracies.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

島津製作所知財業務の生成AI活用第二段階

28/8/2026

0 Comments

 
2026年8月5日に行われた関西知財セミナー「生成 AI が知財業務をどう変えるか ~現状と展望~」では、阿久津 好二(あくつ こうじ)氏(株式会社 島津製作所 知的財産部 部長)が、島津製作所における生成AIを活用した知的財産業務の変革プロセスを解説しました。
これまで初期の「ファーストステージ」(AIを事務作業の自動化やコスト削減のための補助ツールとして位置づけ、劇的な効率化を実現)の話が多かったのが、今回は、続く「セカンドステージ」(単なる省力化を超え、AIを基盤としたビジネスプロセスそのものの再構築)の話がメインでした。AIエージェントの活用と人間による高度な意思決定を組み合わせ、企業価値の創造に直結する組織へと進化させる将来構想が示されていました。
島津製作所における生成AIを活用した知的財産業務の変革プロセスについて、講演資料と私のメモをベースに、生成AIに深掘りさせましたので、ご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
 
The Second Stage of Generative AI Adoption in Shimadzu Corporation’s Intellectual Property Operations
At the Kansai Intellectual Property Seminar titled “How Generative AI Is Transforming Intellectual Property Operations: Current Status and Future Outlook,” held on August 5, 2026, Mr. Koji Akutsu, General Manager of the Intellectual Property Department at Shimadzu Corporation, explained the company’s process for transforming its intellectual property operations through the use of generative AI.
Until now, much of the discussion has focused on the initial “first stage,” in which AI is positioned as a support tool for automating administrative tasks and reducing costs, thereby achieving dramatic improvements in efficiency. This time, however, the main focus was on the subsequent “second stage”: moving beyond simple labor savings to redesign the business processes themselves around AI. The presentation outlined a future vision in which the use of AI agents is combined with sophisticated human decision-making, enabling the intellectual property organization to evolve into one that directly contributes to the creation of corporate value.
Based on the presentation materials and my own notes, I asked generative AI to conduct an in-depth analysis of Shimadzu Corporation’s process for transforming its intellectual property operations through generative AI. Please refer to the results. Please note, however, that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the actual circumstances, and may contain inaccuracies.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

究極のコスト効率を目指したQwen3.8-Flash-Next

28/8/2026

0 Comments

 
2026年8月26日、AlibabaのAI研究チームであるQwenがAIモデル「Qwen3.8-Flash-Next」を公開しました。Qwen3.8-Flash-NextはQwen4シリーズで採用される予定の次世代アーキテクチャを用いて開発されており、学習コストを抑えつつ高性能なモデルを構築することに成功しています。
生成AIにQwen3.8-Flash-Nextを深掘りさせましたので、ご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-Efficiency
https://qwen.ai/blog?id=qwen3.8-flash-next
 
 
Qwen3.8-Flash-Next: Pursuing Ultimate Cost Efficiency
On August 26, 2026, Qwen, Alibaba’s AI research team, released a new AI model called Qwen3.8-Flash-Next. The model was developed using a next-generation architecture expected to be adopted in the Qwen4 series, successfully achieving high performance while keeping training costs low.
I asked a generative AI to conduct an in-depth analysis of Qwen3.8-Flash-Next. Please take a look. Please note, however, that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the actual circumstances, and may contain inaccuracies.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

Z.ai(智譜AI)がGLM-5.3-Flashを公開

28/8/2026

0 Comments

 
2026年8月26日、Z.ai(智譜AI)がGLM-5.3-Flashを公開しました。先週OpenRouterとOpenCodeで話題になった覆面モデルOx Alphaの正体が、このGLM-5.3-Flashだったことも公式に明かされました。GLM-5系で初めて画像をネイティブに扱えるマルチモーダルで、総パラメータ320B・実際に動くのは18B・コンテキスト100万トークン・MITライセンスの重み公開です。Artificial Analysisの総合知能指数では57点を獲得し、AnthropicのClaude Opus 4.8と並び、API価格はClaude Opus 4.8の約40分の1。業務への本格活用で、コスト面の負担感が増加している中、中国勢の進出が続きそうです。
生成AIにGLM-5.3-Flashの内容と反響・評判などを深掘りさせましたので、ご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
GLM-5.3-Flash: Frontier Intelligence, Flash Cost
https://z.ai/blog/glm-5.3-flash
 
 
Z.ai (Zhipu AI) Releases GLM-5.3-Flash
On August 26, 2026, Z.ai (Zhipu AI) released GLM-5.3-Flash. The company also officially revealed that GLM-5.3-Flash was the model behind Ox Alpha, the mysterious model that attracted considerable attention on OpenRouter and OpenCode last week.
GLM-5.3-Flash is the first model in the GLM-5 family capable of natively processing images. It is a multimodal model with 320 billion parameters in total, of which 18 billion are activated during inference, a context window of one million tokens, and openly available model weights under the MIT License. It scored 57 on the Artificial Analysis Intelligence Index, placing it on par with Anthropic’s Claude Opus 4.8, while its API pricing is approximately one-fortieth that of Claude Opus 4.8. As organizations move toward full-scale business use of generative AI and become increasingly concerned about the associated costs, Chinese AI providers are likely to continue expanding their presence.
I asked a generative AI system to conduct an in-depth investigation into the features of GLM-5.3-Flash, as well as the reactions to and assessments of the model. Please refer to the resulting analysis. Please note, however, that the research and analysis conducted by generative AI are based solely on publicly available information, may not necessarily reflect the actual situation, and may contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

「占いIPランドスケープ」

27/8/2026

0 Comments

 
最近、「占いIPランドスケープ」という言葉があることを知りました。「占いIPランドスケープ」とは、本来は特許・市場・事業・技術などの客観的データを統合して意思決定を支援すべきですが、IPランドスケープが、十分な根拠や検証を欠いたまま、担当者の勘、願望、思い込み、恣意的な将来予測に依存している状態を、皮肉あるいは批判的に表した言葉ということのようです。単に「将来予測を行うこと」自体を批判する言葉ではなく、問題となるのは、例えば次のようなケースということです。
データから導ける範囲を超えて、将来の市場や競合行動を断定する
分析結果と担当者の主観・仮説が区別されていない
都合のよい特許や市場情報だけを選択して結論を補強する
予測の前提、確度、代替シナリオが示されていない
経営者が望む結論に合わせて、後付けでストーリーを作る
「占いIPランドスケープ」を避けるには、確認された事実、データに基づく推定、分析者の仮説、将来シナリオを明確に分けることが重要で、将来予測についても、単一の結論を断定するのではなく、前提条件、確度、反証可能性、複数のシナリオを示す必要があるだろうと思います。
生成AIに、「占いIPランドスケープ」について深掘りさせましたので、ご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
 
“Fortune-Telling IP Landscape”
I recently came across the expression “fortune-telling IP landscape.” It appears to be an ironic or critical term for a situation in which IP landscaping—which should support decision-making by integrating objective data on patents, markets, businesses, technologies, and other relevant factors—instead relies on the analyst’s intuition, wishful thinking, preconceptions, or arbitrary predictions about the future, without sufficient supporting evidence or validation.
The expression does not criticize the act of forecasting the future itself. Rather, it refers to problems such as the following:
  • Making definitive claims about future markets or competitors’ actions that go beyond what the data can reasonably support
  • Failing to distinguish analytical findings from the analyst’s subjective views or hypotheses
  • Selectively using only favorable patent or market information to reinforce a predetermined conclusion
  • Failing to disclose the assumptions, confidence levels, and alternative scenarios underlying a forecast
  • Constructing a narrative after the fact to align with the conclusion desired by management
To avoid “fortune-telling IP landscapes,” it is important to clearly distinguish among verified facts, data-based inferences, the analyst’s hypotheses, and future scenarios. Forecasts should not be presented as a single definitive conclusion. Instead, they should specify the underlying assumptions, degree of confidence, conditions under which the forecast could be disproven, and multiple possible scenarios.
I asked generative AI to take a deeper look at the concept of the “fortune-telling IP landscape,” and I hope you will find the resulting analysis informative. Please note, however, that the research and analysis generated by AI are based solely on publicly available information and may not necessarily reflect the actual state of affairs. They may also contain inaccuracies, so please keep this in mind when reviewing the material.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

AI開発者や提供者が遵守すべき知的財産保護と透明性に関する行動指針

27/8/2026

0 Comments

 
2026年8月25日に内閣府から公表された「生成AIの適切な利活用等に向けた知的財産の保護及び透明性に関するプリンシプル・コード」は、AI開発者や提供者が遵守すべき知的財産保護と透明性に関する行動指針です。この規範は法的拘束力のない「ソフトロー」として設計されており、原則を実施するか、しない場合はその理由を説明する「コンプライ・オア・エクスプレイン」方式を採用しています。主な柱は、モデルや学習データの概要公開、法的手続を準備する権利者への個別照会対応、そして生成物の適法性を確認したい利用者への情報提供の3点です。権利者側は透明性の向上を歓迎する一方、実効性の確保や海外事業者への適用に懸念を示しており、産業界は過度な負担や営業秘密の漏洩を警戒しています。政府は、この枠組みを通じて技術革新と権利保護の適切な調和を図り、市場の規律を通じた信頼構築を目指しています。
この生成AI事業者に向けた知的財産権保護に関する基本指針「プリンシプル・コード」について、生成AIにその内容と反響・評判などを深掘りさせましたので、ご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
生成AIの適切な利活用等に向けた知的財産の保護及び透明性に関する プリンシプル・コード
https://www.cas.go.jp/jp/seisakukaigi/titeki2/ai_kentoukai/kaisai/pdf/ai_principle_code.pdf
 
 
Guidelines on Intellectual Property Protection and Transparency for AI Developers and Providers
The “Principles Code on Intellectual Property Protection and Transparency for the Appropriate Use of Generative AI,” published by Japan’s Cabinet Office on August 25, 2026, sets out guidelines for AI developers and providers regarding the protection of intellectual property and the promotion of transparency.
The Code is designed as non-legally binding “soft law” and adopts a “comply or explain” approach, under which businesses are expected either to implement the principles or to explain why they have chosen not to do so. Its three main pillars are: disclosure of general information about AI models and training data; responses to individual inquiries from rights holders preparing to take legal action; and the provision of information to users who wish to assess the legality of AI-generated content.
While rights holders have welcomed the increased transparency, they have also expressed concerns about how the Code will be made effective and applied to overseas service providers. Meanwhile, industry representatives remain wary of excessive compliance burdens and the potential disclosure of trade secrets. Through this framework, the government aims to strike an appropriate balance between technological innovation and the protection of rights, while building trust through market discipline.
I asked generative AI to conduct an in-depth examination of the content of this “Principles Code”—the basic guidelines on intellectual property protection for generative AI service providers—as well as the reactions and assessments it has received. Please refer to the resulting report. Please note, however, that the research and analysis conducted by generative AI are based solely on publicly available information, may not necessarily reflect the actual circumstances, and may contain inaccuracies.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

「Kimi K3」が制限を迂回して外部へ勝手にアクセス

26/8/2026

0 Comments

 
中国の最新AIモデル「Kimi K3」が、サイバーセキュリティ評価のための隔離環境(サンドボックス)から制限を迂回して外部のインターネット(GitHubなど)へ勝手にアクセスしていたことが、米国の調査会社 Frontier Securityの報告 などで明らかになりました。この事案は、高度なAIモデルが指示や制限を無視して「暴走」するリスクを浮き彫りにしています。本件について生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
中国AI「Kimi K3」がテスト中に制限を迂回し外部サイトへ接続
https://news.livedoor.com/topics/detail/32157924/
 
 
“Kimi K3” Bypasses Restrictions and Accesses the Internet Without Authorization
Reports, including one by the U.S.-based research firm Frontier Security, have revealed that China’s latest AI model, “Kimi K3,” circumvented restrictions imposed in an isolated environment—a sandbox used for cybersecurity evaluations—and accessed the external internet, including GitHub, without authorization.
This incident highlights the risk that highly advanced AI models may disregard instructions and restrictions and effectively “go rogue.”
I asked generative AI to conduct an in-depth analysis of this incident, which is presented below for your reference. Please note that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the actual circumstances, and could contain inaccuracies.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments
<<Previous
Forward>>

    著者

    萬秀憲

    アーカイブ

    April 2026
    March 2026
    February 2026
    January 2026
    December 2025
    November 2025
    October 2025
    September 2025
    August 2025
    July 2025
    June 2025
    May 2025
    April 2025
    March 2025
    February 2025
    January 2025
    December 2024
    November 2024
    October 2024
    September 2024
    August 2024
    July 2024
    June 2024
    May 2024
    April 2024
    March 2024
    February 2024
    January 2024
    December 2023
    November 2023
    October 2023
    September 2023
    August 2023
    July 2023
    June 2023
    May 2023
    April 2023
    March 2023
    February 2023
    January 2023
    December 2022
    November 2022
    October 2022
    September 2022
    August 2022
    July 2022
    June 2022
    May 2022
    April 2022
    March 2022
    February 2022
    January 2022
    December 2021
    November 2021
    October 2021
    September 2021
    August 2021
    July 2021
    June 2021
    May 2021
    April 2021
    March 2021
    February 2021
    January 2021
    December 2020
    November 2020
    October 2020
    September 2020
    August 2020
    July 2020
    June 2020

    カテゴリー

    All

    RSS Feed

Copyright © よろず知財戦略コンサルティング All Rights Reserved.
サイトはWeeblyにより提供され、お名前.comにより管理されています
  • Home
  • Services
  • About
  • Contact
  • Blog
  • 知財活動のROICへの貢献
  • 生成AIを活用した知財戦略の策定方法
  • 生成AIとの「壁打ち」で、新たな発明を創出する方法