ART
Practical intelligence for global IP professionals. Published monthly by Park IP.
July felt like the month the IP world moved from debating whether AI changes patent practice to arguing about how. Japan tightened its inventive step standard for AI-assisted applications. The USPTO’s foreign representation rule took effect today: a structural shift that arrives without fanfare but reshapes thousands of workflows overnight. And the world’s five largest patent offices are now formally aligned on AI examination, with a dedicated working group established. The question “where does automation end and accountability begin” is no longer hypothetical. This issue takes a position.
Three stories. Everything you need to know this month.
The JPO’s Status Report 2026 (March 2026) tightened the inventive step standard for AI-assisted inventions. Examiners are now directed to determine whether a claimed invention reflects a genuine technical contribution beyond what a person skilled in the art, aided by commonly available AI tools, could independently achieve. The March 2026 examination handbook revision added new case examples covering generative AI applications and AI estimation methods. Voluntary disclosure of AI involvement in prosecution is now encouraged; a mandatory requirement is expected to follow.
Applications drafted without explicitly articulating human-directed design choices are now more exposed at JPO. If your team is prosecuting AI-related applications in Japan (or planning to), the specification language needs to be reviewed against the new standard before your next office action lands. Pendency is 9.1 months on standard examination. You won’t get a second chance to reframe before scrutiny arrives.
The USPTO’s final rule requiring all foreign-domiciled patent applicants and patent owners to be represented by a registered U.S. practitioner took effect July 20, 2026. The rule applies to new filings, pending applications, and existing patent matters. Approximately 97% of foreign-origin non-provisionals already list a U.S. practitioner, but the remaining 3%, plus any applicant whose U.S. counsel has changed since the application was filed, face immediate exposure.
This is not a future risk. It is a today risk. Any foreign-domiciled applicant or patent owner without a listed U.S. practitioner of record is currently non-compliant. Review your portfolio for any affected matters, particularly applications filed directly, PCT nationals without U.S. counsel, or matters where previous counsel has changed.
At the 19th IP5 Heads of Office Meeting on June 12, 2026, hosted by the JPO in Tokyo, patent chiefs from Japan, Europe, South Korea, China, and the United States reviewed progress on the IP5 NET/AI Roadmap and agreed on new directions for AI cooperation. The five offices formally established a dedicated AI working group tasked with advancing practical discussions and coordinated initiatives across the IP5 system. The day prior, the offices convened with industry representatives to exchange views on AI’s current status and prospective use in examination.
Five offices accounting for the majority of global patent filings are now formally structured around AI coordination, not aspirationally but with a working group mandate and a signed joint statement. For IP teams prosecuting across these jurisdictions, the question is no longer whether AI examination standards will converge, but at what pace. The working group’s output will define what “aligned AI examination” actually means in practice, and it will shape prosecution strategy across every major market.
Three things worth two minutes of your time.
CLAIMS
One debate. Every issue. No easy answers.
Modern neural machine translation has reached quality thresholds that make full review unnecessary for routine, lower-stakes applications. For high-volume portfolios filing in multiple languages, mandatory review of every translation creates bottlenecks and cost overhead that slow the entire prosecution timeline. Applied selectively (with AI review confined to mature translation pairs and technically straightforward descriptions), the risk is manageable and the efficiency gains are real.
Translation error in a patent claim creates scope problems that may not surface until litigation; at which point correction is expensive, time-consuming, and in many jurisdictions, not permitted after grant. The asymmetry matters: the cost of review is predictable; the cost of a narrowed claim is not. There is no such thing as a low-stakes translation when claim scope is on the line.
Both sides agree AI translation quality has improved. The disagreement is about accountability. When an AI-generated translation is filed and later found to have shifted claim scope, who owns that outcome: the AI vendor, the filing firm, or the client? Until that question has a clear contractual and legal answer (and right now, it doesn’t), the risk calculus favours human review. Every time.
Where does your team stand? Share your take on LinkedIn →
Three things that aren’t news yet, but will be.
Legal Studio puts accurate cost data in front of IP teams before the filing window closes. Most PCT cost decisions happen in the final two weeks of the 30-month window, when deadline pressure makes optimization functionally impossible. Run a full multi-jurisdiction PCT estimate in four minutes. For EP, get translation and official filing fees broken down by designated state (Germany, France, the UK, and beyond) before the validation deadline arrives. With the UPC reshaping European filing strategy, knowing the cost differential between validation paths has become a filing decision, not just a budget line. Cost visibility doesn’t just improve budgeting: it changes the jurisdictions you file in, which changes where your IP is protected and at what cost.
Try Legal Studio →That’s 23% generative AI and 9% agentic AI. Agentic alone has grown 59% over the past two years. The volume of AI-related prosecution is no longer something to monitor. It’s a workflow challenge to solve.