What Are the Risks of Generative AI for the Patent Law Profession?

For many legal professionals, artificial intelligence platforms are being adopted at a speed that they think is imprudent. Well-respected patent and intellectual property thought leaders have been very open about their concerns on this point. From client confidentiality to patentability, risks posed by the use of generative AI systems must be eliminated by AI companies themselves before they partner with the patent law profession.

Intellectual property law is among the practice areas expected by practitioners to see the most transformation due to generative AI use, according to an industry survey issued last August. This survey also revealed that two-thirds of all legal professionals believe artificial intelligence will be critical to law firm success within five years.

The use of generative AI could eliminate hours of patent application drafting work seems a natural conclusion, but several concerns about AI’s impact on patent law must be addressed head-on by any tech vendor partnering with this sector.

1. Security and Confidentiality

Patent law is unforgiving on parties who do not maintain secrecy before filing a patent application. The first public use or sale of an invention begins a 12-month window during which an inventor must file a patent application in several major nations. Other countries only provide a six-month window between disclosure and filing. Inadvertent leaks of inventive subject matter can destroy an invention’s novelty, which is critical to patent eligibility and cannot be risked.

One of the biggest obstacles preventing the legal profession from employing generative AI is the use of confidential client information in AI training models. This puts the needs of the generative AI platform ahead of the legal profession, which is completely backward. In the patent world, first-to-file rules can leave an authentic inventor empty-handed if a wealthy, unscrupulous competitor gains access to invention disclosures.

Generative AI companies that are serious about answering the data security needs of the patent profession must prove that their architecture has no cracks. Zero data retention and ethical walls are safeguards AI platforms can implement to protect valuable information from prying eyes. Coupling those protective measures with an on-premises approach to training AI models that complies with information security standards can sandbox data within a patent attorney’s own computer systems.

2. Trustworthiness

Attorneys everywhere have heard horror stories about fictional cases hallucinated by AI systems while drafting legal briefs. In the United States, federal judges have begun requiring attorneys to identify portions of legal filings drafted by generative AI systems. Inaccurate legal briefs threaten the integrity of the judiciary and waste a court’s time trying to separate fact from fiction. For patent attorneys, patent applications with improperly worded claims or specifications can give rise to costly enablement or indefiniteness issues when bringing an infringement suit.

Having a human in the loop to closely review generative AI outputs is critical, especially today during the advent of generative AI. It is the responsibility of the AI provider to understand the issues presented by inaccurate claim limitations. Patent agents will not trust tech providers who allow those agents to walk blindly into hallucinations, even with the benefit of hours of extra time to review drafted patent applications.

While attorney review can never be entirely eliminated by AI, generative AI platforms can employ adversarial robustness techniques to prevent the generation of text that is clearly wrong but looks deceptively correct. These techniques teach AI models how to handle noise and data corruption during the training process, making those platforms more resistant to data irregularities.

3. Ethics and Compliance

In addition to industry-wide model rules of ethics that every lawyer must follow, patent attorneys must adhere to rules of conduct and duties of candor imposed by filing offices. U.S. Patent and Trademark Office (USPTO) Director Kathi Vidal has recently provided clarification on the application of current Rules of Professional Conduct to AI.

Several IP agencies are currently developing their approaches to generative AI. While inventorship issues might be the chief concern for these agencies, many patent attorneys are well aware that inadvertent disclosures of privileged information by AI platforms create an impermissible risk of malpractice liability.

The legal profession’s adoption of cloud computing technologies over the past decade provides a useful case study. In the United States alone, the percentage of lawyers using cloud platforms for practice or document management jumped from 21% in 2012 to 70% in 2022. Considering many of the same ethical concerns posed by AI, 30 state bar associations have issued guidance informing member attorneys how to adopt cloud computing tools while complying with rules of conduct.

Then there are ethical concerns about the use of generative AI that aren’t covered by a lawyer’s duty of competent representation. Training AI models with legally protected content has led to infringement lawsuits in the copyright context. Vendors marketing AI-powered drafting solutions for patent agents and attorneys need to develop and maintain data practices that encourage good behaviors among all participants.

4. Regulation

Patent lawyers understandably want to avoid the confusion created by a flurry of regulatory activity surrounding AI platforms. First introduced in 2021, the European Union’s AI Act was the focus of serious negotiations among lawmakers this past summer, leading to a provisional agreement in December 2023 on the AI Act’s framework for regulating high-risk AI applications. In the United States, President Joe Biden issued an executive order in late October directing government agencies to address unfair competition, civil rights and other concerns in the nascent AI economy.

Generative AI platforms have already been the subject of scrutiny for data protection authorities in several countries. Privacy regulators in several countries, including the United States, Italy and South Korea, have opened investigations into the data collection practices of ChatGPT. The results of these investigations, and any enforcement actions, will serve as early rules of the road for the generative AI market.

National regulatory authorities are issuing guidelines that AI companies can implement to protect personal information while improving the accuracy of artificial intelligence models. One technique that addresses regulator concerns is differential privacy, which creates noise in datasets designed to anonymize sensitive personal information. Generative AI companies must demonstrate that they can deftly handle regulatory challenges with such techniques if they hope to partner with professionals in the high-stakes world of patent law.

The Road to Harmony Will Be Bumpy

In conclusion, navigating the intricate landscape of AI-driven patent drafting involves addressing multifaceted challenges across security, trustworthiness, ethics, and compliance, as well as staying abreast of evolving regulatory frameworks. The imperative to safeguard confidentiality and navigate legal intricacies demands robust security measures, and the integration of human oversight becomes pivotal to ensure trustworthiness and accuracy. Ethical considerations and compliance with evolving regulations underscore the need for responsible AI practices. As the legal tech sphere evolves, a strategic and collaborative approach is essential, with a commitment to staying ahead of the regulatory curve and fostering ethical AI deployment in the patent law domain. The journey toward a harmonious collaboration between legal practitioners and AI necessitates a nuanced understanding of these challenges and a proactive stance toward innovation within established ethical and legal boundaries.

Image Source: Deposit Photos
Author: Feodora52
Image ID: 325430688