Offensive And Defensive IP For AI In Health Care

Offensive And Defensive IP For AI In Health Care

January 9, 2023 - Artificial Intelligence ("AI") continues to grow in importance, especially in healthcare and health technology. AI can help marketers, technology companies and researchers better understand a health condition or disease in less time, and the value of AI's predictive capabilities will only increase with more training data.

Protecting intellectual property ("IP") involved in products or services that use artificial intelligence is important for companies to maintain a competitive advantage. A good IP strategy not only protects your product from competitive marketing and legal challenges, but also puts you in a good position to protect your brand and market share.

There are various IP strategies that can protect the underlying products and data. This article examines the most relevant intellectual property protections that businesses may face: patents, copyrights, and trade secrets, the existence of intellectual property, and the advantages and disadvantages of each.

Proprietary software upon registration with the United States Patent and Trademark Office

The license granted allows you to prevent others from making, using, selling or reselling it. In the case of AI-powered products, several elements may be patented, namely the AI ​​software, the AI ​​data, the information, and the overall product analysis process. Unfortunately, patent protection is not available for all of these features under current US law, but that hasn't stopped companies from trying.

In 2020, the United States Patent and Trademark Office ("USPTO") reported that patent applications for AI software more than doubled between 2002 and 2018. AIs must be new and innovative to be patented. In 2020, the USPTO received nearly 80,000 AI software patent applications and approved 77% of them.

Patents may be granted for, inter alia, inventive utilities or processes. For AI-powered products, data is not proprietary in itself, but may be protected by other intellectual property rights, as we explain below. The same goes for any valid code. But other things, like how the AI ​​works, how the AI ​​uses or integrates data, and the product itself can be patented. US law places certain restrictions on the acquisition of patents related to artificial intelligence products.

The United States Supreme Court's 2014 decision in Alice Corp. v. CLS Bank does not patent the implementation of "abstract ideas" in computers; He said it casts a shadow over software patents in general. Alice has created a two-part patent eligibility test for patent examiners and courts to determine whether AI-related patent applications are eligible for patents. In a December 2021 trial, a federal district court in Texas (Health Discovery Corp. v. Intel Corp.) ruled that machine learning ("ML") technology patents were invalid because they covered abstracts of ideas that did not pass the Alice test.

Given the uncertainty surrounding the patenting of AI-related products, it is important to consider whether seeking patent protection is the best strategic option given the potential downsides of seeking patent protection. Public disclosure is a big problem. Your patent application should provide enough information to describe your invention, which usually includes a detailed description of your products and processes.

When you file a patent application, it becomes public 18 months after filing the application. Once published, anyone, including your competitors, can use this information. If you never get a patent, it means you know how your product works without using any means to prevent others from using this information. The current time from patent filing to grant is two to three years, which means it allows you to go public without knowing whether you will receive a patent or not.

Patent protection lasts for 20 years, so it is ultimately a very powerful right, but before you file a patent application, you should think carefully and think about the issues you need to apply for patent protection. .

Copyrighted AI software

Copyright is another form of IP protection that should be included in your IP strategy, as AI software can be granted rights to respect patents and trade secrets. Copyright protects original human-made works. Earlier this year, the US Copyright Office announced that the copyright group generally does not grant copyright to original works created by artificial intelligence because AI does not consider itself an author, a human being.

For AI software, you can look for copyright protection and copyrights for the internal processes of the AI ​​software, such as training data. Copyright applications related to AI software are usually filled with training information to protect this selection and collection. Training information should not only be man-made but also original.

The USPTO generally considers training data to be original if there is something unique about the selection, inclusion, and/or presentation of the AI ​​software data. Therefore, copyright may not protect artificial intelligence software that helps diagnose a disease, but it may protect selected training data for a particular disease.

Copyright should belong to your organization, not individual developers or employees. To ensure that your business has the right to enforce copyright, your company must properly assign rights to all participants in the creation of the material, including employees and contractors. Additionally, your copyright protection strategy should include a nondisclosure and/or confidentiality agreement with your AI software development team and any third parties for your training data. Such agreements prohibit their developers or third parties from disclosing the AI ​​software and the training data used for the AI ​​software.

Keep everything secret. trade secret

IP security is an important trade secret that is sometimes overlooked by developers and technology companies. Although specific requirements vary by jurisdiction, a trade secret is generally confidential information of commercial value that is kept secret for the economic benefit of the owner. Trade secrets can protect training information, as well as any process, method, procedure or other sensitive information under AI itself. Trade secrets may include, for example, algorithms, ML, research and development information, health information, and business information related to AI software.

But unlike a copyright or patent, a trade secret is a defense strategy because the information you want to protect should be kept private and not in the public domain. By its very nature, unlike a patent, a trade secret does not alert competitors to your intellectual property.

However, a trade secret can be an important part of your IP strategy, especially in some cases as an alternative to patent or patent protection that cannot be secured. When your AI software fails the Alice test, or if the costs of seeking patent protection outweigh the benefits, intellectual property protection may be a trade secret option.

To protect trade secrets, a company should consider appropriate internal policies that ensure continued confidentiality, including by limiting access to artificial intelligence software and training data. These policies should consider the distribution of remote access and remote work situations. All persons with access to trade secret information must enter into appropriate confidentiality agreements, and no trade secret must be disclosed to third parties without a satisfactory non-disclosure agreement.

However, efforts to protect information such as trade secrets can create a "black box" problem, preventing information sharing, which is especially important for information and technology-based health technology innovations in the health information complex. Moreover, even if a trade secret is protected for the necessary period, it is conceivable that a competitor could provide artificial intelligence software or similar training data before the period expires. Trade secret protection does not protect against a competitor's use of your own independently developed software or data. Thus, a new IP strategy for healthcare AI software can use trade secrets along with the other two IP protections.

Resume:

There is no uniform IP security policy for AI software. Your intellectual property protection strategy should consider your business goals as well as the benefits and limitations of proprietary rights, copyrights, and trade secrets. A successful IP security strategy can include a mix of offensive and defensive IP security.

The opinions expressed are those of the author. They do not reflect the views of Reuters News, which is committed to honesty, independence and impartiality based on the principles of integrity. Westlaw Today is owned by Thomson Reuters and operates independently of Reuters News.

Jason Johnson

Jason Johnson is a strong partner in the Healthcare, Privacy & Cybersecurity, and Intellectual Property practice groups. His practice focuses on the legal aspects of digital health innovations, data privacy and security under US and European law, as well as complex regulatory and compliance issues related to clinical research and corporate affairs. His clients include academic medical centers, health technology companies, biotech start-ups, pharmaceutical and medical device companies, and other research and healthcare organizations. He can be reached at JJohnson@mosessinger.com.

Blaze D. Valeski

Blaze D. Valeski is a firm consultant who spends much of his time on privacy, data protection and technology, particularly in the financial and healthcare industries. He can be reached at bwaleski@mosessinger.com.

Learn Caro-Kan Defense | 10 minutes of open chess

Tidak ada komentar untuk "Offensive And Defensive IP For AI In Health Care"