Healthcare has seen a transformation in the past decade. Digitized medical charts and patient web portals have changed the face of day-to-day medical practice, and we shouldn’t expect the changes to slow down any time soon; the global trend of increased focus and reliance on tech shows no signs of reversing.

The inroads being made in computational neuroscience are some of the most exciting applications of Artificial Intelligence; researchers are using neural networks — essentially computerized models of a brain capable of machine learning- in order to better understand complex brain signals and treat a variety of neurodegenerative ailments, in addition to Post Concussion Syndrome and CTE.

In an article in the Financial Times, Benjamin Fels lays out some of the major advantages of adopting machine learning technologies in medicine, principally that the ability of AI to appreciate degrees of similarity and difference between individuals will enable implementation of high-quality, tailored treatments on a scale unfathomable to the current medical world. Companies are already using lab testing data as well as biological information gathered directly from patients to shake up the traditional top-down model of drug and treatment development. This is great news, as current procedures too often have significant blindspots that lead to poor patient outcomes due to the limitations of conventional data collection and testing. Investors are banking on this work producing valuable innovations; the health startups on the forefront of this trend are receiving millions of dollars in funding every year. All of this data collection makes the future of individualized treatment and healthcare development look so promising that it begs the quesion: what’s the catch?

The introduction of AI does signal a hopeful future for the treatment of disease, but there are potential dangers as well. Primary among them is the issue of a biased dataset. Any algorithm or AI utility relies on the initial supply of data it’s given to guide its work, and in a study published in the medical journal JAMA Internal Medicine the coauthors express their concerns that poor data collection practices could exacerbate the pre existing inequalities of access to medical care and lead to deadly consequences. Larger, more general issues of privacy and data security have been a familiar concern of the medical community for years, and with discussion of using crowdsourced cell phone data to drive medical analytics already underway, it’s critical to take steps to ensure that private and confidential data is handled legally and responsibly.

Benefits and pitfalls notwithstanding, AI and machine learning is coming to medical tech and it’s unlikely that we’ll see a reversal in this trend. It is therefore imperative that any organization entering this field seek and follow guidance from governmental regulatory and research organizations like the NIH and the CDC, as well as from legal experts and relevant, contemporary scholarship to ensure that their products follow through on their intentions of improving and, importantly, democratizing access to quality treatments. If the entities involved in developing these technologies take seriously their obligation to all patients and communities that stand to benefit from them, AI and machine learning have the potential to usher in a new era of improved medical intervention.