Kilpatrick’s Charles Gray and Karam J. Saab recently presented on the topic of “AI in Practice: Patent Prosecution Strategies and the Upcoming Colorado AI Mandates” at the firm’s annual “SKI”-LE held in Breckenridge, Colorado. Generative AI represents a paradigm shift for patent practice, but it requires a calibrated approach. Attorneys must distinguish between where AI succeeds (technical descriptions, claim charting, and summarization) and where it fails (authoritative legal research and hallucination risks). This session defined the safe use cases for AI in patent prosecution and explored the ethical boundaries from both attorney and client perspectives. With the Colorado Artificial Intelligence Act (CAIA) taking effect on June 30, 2026, the regulatory landscape is shifting. 

Key takeaways from the presentation include:

1. Prepare for the Colorado AI Act (CAIA)
The CAIA is to be the first state law to comprehensively regulate AI systems. Slated to take effect June 30, 2026, it applies to parties doing business in Colorado developing or deploying AI systems that make “consequential decisions” related to the procurement or cost of employment, financial services, healthcare, housing, insurance, and legal services for consumers.

• A primary goal is eliminating “algorithmic discrimination” resulting in the unlawful differential treatment of an actual or perceived protected class.
• Imposes specific duties of care and impact assessment requirements on both Developers and Deployers.
• Consumers will have the right to appeal decisions and correct any inaccurate personal data used by the AI to reach an adverse decision.

2. Master AI Prompt Engineering
Overbroad prompts lead to errors, while clear and specific instructions equal optimal success. To help encourage use, firms should establish a shared prompt library to facilitate new users being able to effectively use an AI chatbot.

• Provide specific examples in the exact format and style you want as part of your prompt.
• Ask the AI engine for feedback on how to reformulate your prompts to achieve better results.

3. Clients can use AI as a starting point to target investment and assess attorney work product
AI works particularly well within a defined universe of technical documents. A couple use cases that have already made significant differences in how our clients manage patent prosecution include:

• Use AI to compare new invention disclosures with the client’s existing patent portfolio to assess possible overlap. This function is particularly useful in circumstances where multiple in-house counsel manage a portfolio.
• Use AI to compare an invention disclosure and transcript of an invention disclosure meeting with a drafted patent application to ensure that all concepts considered by the inventor have been described within the patent application.

4. Enforce Strict Ethical Guardrails
AI tools carry inherent risks that must be managed responsibly. As the attorney of record, you own your final work product.

• Confidentiality: Never input client-privileged information into public AI platforms. Rather, always use secure, enterprise-grade instances.
• Hallucinations: AI models can fabricate information. Since May 2025, judges have cited at least 23 instances of AI hallucinations in court records. Independently verify all output.