Practitioners who build durable AI patent portfolios start by deeply understanding the invention—then strategically apply USPTO guidance to position it effectively.

At IPWatchdog LIVE 2026, the Day 3 panel "AI Patents Under the Microscope: Drafting and Prosecuting for the Long Game" cut through the noise. Participants from Salesforce, Capital One, Ballard Spahr, Blueshift IP, and Ankar AI emphasized that the strategies that worked 18 months ago—amid the February 2024 AI inventorship guidance—may now actively undermine an application filed today, especially after the USPTO's November 2025 rescission returning to traditional human-conception standards.

Takeaways from the Panel

  • ⚖️ The guidance-to-practice gap runs both ways. New memos do not instantly reshape examiner behavior. As Jason Harrier (Salesforce) asked, "How long does it take for guidance to trickle down?" Practitioners citing the latest memo may still face examiners operating on prior instructions.
  • 🎛️ Use continuations strategically. Robert Plotkin (Blueshift IP) files the narrowest, most defensible claims first and pursues broader claims through continuations. An early allowance changes the examiner's orientation toward the entire portfolio.
  • 📖 The invention narrative is the only durable foundation. Thara Russell (Capital One): "Trying to shoehorn claims to align with the most recent guidance is a recipe for headache." Start with what the inventor actually built and why it improves the prior art. Build layered fallback positions in the specification; use guidance only to position prosecution where the examiner deviates from it.
  • 🤖 AI drafting tools are useful but not without risks. The invention story is not delegable. As Giulia Toti (Ankar AI) warned, AI-generated output that "looks accurate" is no substitute for a specification grounded in what the inventor actually built; hallucinations become liabilities in examination or litigation. Inventor and attorney review remains critical.
  • 🚫 Sometimes the right advice is not to file. If the achievable claim scope is too narrow to be commercially meaningful or defensible post-grant, a filed application can be a cost sink. Assess the full range of options against the client's objectives, including trade secret protection for features that are difficult to reverse engineer.

In the current AI Gold Rush—where generative tools produce 40-page disclosures instead of two-sentence descriptions—this invention-first philosophy is more critical than ever.

Full article by Tim Bright via IPWatchdog, Inc: Read the full article