{"rewrite":{"id":"r_54181809beeaa2580665bc12","clusterId":"c_2fbda1ff19d6b38fe682dcb6","slug":"japan-patent-office-briefs-vcs-on-ai-assisted-ip-due-diligence","model":"deepseek-v4-flash","headline":"Japan Patent Office Briefs VCs on AI-Assisted IP Due Diligence","summary":"The Japan Patent Office held the first briefing for the 2026 round of its VC-IPAS program on May 15, in Tokyo and online. Speakers walked through the program's IP due diligence manual, which splits investment review into a preliminary screening stage and a full DD stage, and demonstrated how generative AI can handle early analysis. The sessions also flagged the limits: hallucinated output, confidential data handling, and prompts that need clear purpose, procedure, definition, and output format.","whyItMatters":"The JPO is positioning generative AI as the low-cost entry point that lets VCs without IP expertise run preliminary screening themselves before spending on outside specialists.","webCardHtml":"\u003cp\u003eThe manual cited throughout the session was compiled from last year\u0026#39;s program, and JPO startup support team head Kazuya Minato said it lowers the hurdle for busy VCs by spelling out how to question startups and commission experts. Patent attorney Takuya Ishii split the process into a preliminary stage with four checkpoints and a full DD stage with six, using a fictional robot that harvests aeroponically grown melons to test whether a patent covers the business. The AI sessions followed the same logic: Akihisa Tsuchimoto proposed prompts built from purpose, procedure, definition, and output format that work from public information alone, while Yoshihide Kakubuchi warned that owning a patent does not prevent infringing someone else\u0026#39;s, which is why claim charts matter.\u003c/p\u003e","blueskyPost":"The JPO's VC-IPAS briefing walked VCs through a two-stage IP due diligence flow and showed how generative AI can handle the early screening. The pitch: structured prompts built from public info, with experts reserved for the deep dive. Hallucination and confidential data remain the watch items.","twitterPost":"JPO's VC-IPAS briefing: two-stage IP due diligence for VCs, with generative AI doing the early screening via structured prompts from public info. Experts stay for the deep dive. Hallucination and confidential data are the flagged risks.","threadsPost":null,"newsletterBlurb":"The Japan Patent Office's VC-IPAS program held its first briefing for the 2026 round, walking VCs through a two-stage IP due diligence flow. Speakers showed how generative AI can run the preliminary screening from public information alone, and where it needs guardrails: hallucination, confidential data, and prompt structure.","attributionJson":"[{\"source\":\"ASCII.jp\",\"url\":\"https://ascii.jp/elem/000/004/428/4428807/?rss\",\"title\":\"「知財」が成否を分ける？──生成AIを活用したDD効率化の実務とは？\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":6290,"outputTokens":2793,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1788163407,"createdAt":"2026-08-31T07:56:40.000Z","publishedAt":"2026-08-31T07:57:27.000Z","updatedAt":"2026-08-31T07:57:27.000Z"},"cluster":{"id":"c_2fbda1ff19d6b38fe682dcb6","canonicalTitle":"「知財」が成否を分ける？──生成AIを活用したDD効率化の実務とは？","representativeArticleId":"a_3dca8183a26bf9406c929b24","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[],\"people\":[],\"type\":\"event\",\"domain\":\"industry\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-08-31T07:00:00.000Z","lastSeenAt":"2026-08-31T07:00:00.000Z","updatedAt":"2026-08-31T07:57:27.000Z"},"attribution":[{"source":"ASCII.jp","url":"https://ascii.jp/elem/000/004/428/4428807/?rss","title":"「知財」が成否を分ける？──生成AIを活用したDD効率化の実務とは？"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":[],"people":[],"type":"event","domain":"industry","is_roundup":false},"keyFacts":null}