Product liability is always a point of concern for manufacturers of medical devices and other U.S. FDA-regulated products, and the broad contours of product liability jurisprudence are well known by corporate counsel. However, artificial intelligence products are rapidly pressing their way into routine clinical use, representing a technological shift that may occasionally deviate from the existing rules of the road where product liability is concerned.
The U.S. FDA’s device center is working to refine its regulation of artificial intelligence algorithms, but the agency is recommending that industry be more forward-thinking in a blog that urges device makers to fully adopt a life cycle management mindset for these systems.
For the third time in as many years, Health Canada, the U.S. FDA and the UK Medicines and Health Care Products Regulatory Agency have teamed up to issue a set of recommendations for artificial intelligence used in or as a medical device.
The U.S. FDA is keen to develop tools for oversight of artificial intelligence (AI) as demonstrated by a batch of research projects designed to inform the review of medical applications of AI. The agency’s concern is that there is a dearth of “robust evaluation methods” for evaluating AI products, thus the need for tools that will allow the agency to clear or approve such products with an assurance that these algorithms are safe and effective for their intended use.
The U.S. FDA might still be seen as the premier med tech regulatory entity in the world, but the agency is badly outnumbered by companies in the life sciences, which are pumping out artificial intelligence algorithms at a breathtaking pace. Further, the FDA must also avoid being lapped by industry in connection with the regulatory novelty known as the predetermined change control plan, a challenge that put the agency’s device center in scramble mode for essentially the entirety of calendar year 2023.
With one program in the clinic and another not far behind, Generate Biomedicines Inc. raised $273 million in a series C financing to advance its generative biology platform. It is one of the largest venture capital (VC) rounds for a U.S. company in 2023. Funds will go toward advancing the Somerville, Mass.-based company’s 17 pipeline programs, including the filing of multiple IND applications in 2024.
The EU’s Artificial Intelligence (AI) Act is still in the thick of the legislative process, which seems likely to ladle even more regulatory liabilities onto AI software used for medical purposes. Bodo Wiegand, senior advisory at Viopsy, told attendees at a May 18 webinar that between the promise of yet more regulation along with existing coverage and reimbursement hurdles in the EU, developers of medical software are considering whether they should steer clear of developments that qualify as AI simply because of the extraordinary time and expense associated with generating revenues for these projects.
Developers of artificial intelligence (AI) and machine learning (ML) algorithms have found themselves returning repeatedly to the U.S. FDA for seemingly modest updates to their products, a problem that may soon be relieved by an FDA draft guidance on predetermined change control for AI and ML. However, Brad Thompson of Epstein, Becker & Green, P.C., told BioWorld that the terms of the draft “hugely increases the burden on developers to plan ahead” in order to obtain that postmarket relief from repeated 510(k) filings, a concession that device manufacturers and software developers may be more than willing to make.
The Biden administration has released a blueprint for an artificial intelligence bill of rights, which is accompanied by an acknowledgement that these algorithms can be crucial in guiding treatment of cancer patients.
The most conspicuous part of the data problem for artificial intelligence (AI) medical software is the bias problem, but the U.S. Government Accountability Office (GAO) says there are policy solutions despite the data ownership/monetization problem.
