130 THE US-ISRAEL | Legal Review 2025/26 be non-transparent. Opponents may demand discovery on prompts, outputs, and internal AI policies, raising confidentiality and privilege issues. The uncertainty itself is risk. When the ability to prove a clean story is impaired, settlement values shift, and enforcement becomes harder. Best Practices for Patent Professionals: A Practitioner Checklist A disciplined AI policy should distinguish safe from risky use cases and should be implementable without halting innovation. Safe(ish) uses » Summarizing non-confidential public prior art. » Translating public documents (with human review for nuance). » Formatting, grammar, and style improvements on redacted drafts. » Generating administrative checklists, timelines, and filing reminders. Higher-Risk Uses That Require Controls » Drafting claims based on confidential disclosures. » Generating embodiments or alternatives not validated by inventors. » Patentability opinions on unpublished inventions. » Searching for “solutions” to technical problems during development. If such uses are permitted, firms should adopt safeguards: enterprise tools with contractual confidentiality commitments, disabling training on user data where possible, strict redaction, internal logging, and mandatory human verification. Documentation Practices » Create a dated human invention disclosure before AI interaction. » Record what prompts were used and what outputs were generated (privately). » Maintain a clear separation between public-art analysis and confidential invention development. » Document tests and experimental results supporting disclosed embodiments. » Confirm that AI-generated technical assertions are supported by inventor knowledge. Client Counseling Clients should be advised—early—that casual AI use can jeopardize global patent rights. The counseling should emphasize that the US grace period does not save European rights, and that AI-mediated disclosure may be difficult to detect until it is too late. Conclusion AI will remain a powerful force in patent practice. The goal is not to prohibit its use, but to prevent careless use from undermining patent rights. The most dangerous mistake is to anthropomorphize AI—to treat it as a trusted colleague rather than as a probabilistic system optimized for producing helpful outputs. If practitioners treat AI as a good but talkative assistant, limit it to safe tasks, and document human conception rigorously, they can gain efficiency without sacrificing rights. Otherwise, the bear hug may tighten: first as a time-saver, then as a novelty problem, and finally as an enforceability risk. Professional Responsibility, Privilege, and Cross-Border Practicalities Beyond patentability doctrine, AI use intersects with professional responsibility and privilege in ways that matter in real disputes. Accordingly, AI governance
RkJQdWJsaXNoZXIy MjgzNzA=