As industries around the globe wrestle with where and how often artificial intelligence (AI) should be implemented in their workflows, biopharma, for one, is proceeding with caution.
In a recent ZoomRx survey of more than 200 life sciences professionals, more than half said their companies have banned employees from using OpenAI’s popular generative AI tool ChatGPT, including 65% of the top 20 Big Pharmas. Respondents said those policies were largely linked to concerns that sensitive internal data could be leaked to competitors.
“The question comes down to: How do you balance the tradeoff between the speculative and the empirical; the potential business benefits vs the recognized security risks? And I think for many of these life sciences companies, particularly those in the top 20, the repercussions of any data mishap currently outweigh any of the perceived outcomes,” Andrew Yukawa, a ZoomRx product manager, said in a statement to Fierce Pharma Marketing.
Yukawa referenced a bug that required a full ChatGPT shutdown last year to fix and that was allowing some users to see other users’ chat history, adding, “You can imagine a scenario where an employee uploads proprietary clinical data or sensitive information from a conversation they had with a colleague into ChatGPT when seeking help at work, and that particular action then inadvertently adds the information into OpenAI’s training data set used to improve future performance of its chatbot.”
Prevention efforts beyond outright bans, however, are lacking: The survey found that fewer than 60% of life sciences companies have provided employees with any kind of training or guidelines about safe use of ChatGPT, though another 15% said they were planning to.
Even so, while about half the industry has banned the tool—and even more companies have yet to conduct any related training—many life sciences professionals are using ChatGPT regularly. More than half of those surveyed said they use it at least a few times a month, and over a quarter of respondents are using the tool a few times a week, if not every day.
When it comes to AI as a whole, though a whopping 83% of those surveyed labeled the technology “overrated,” many life sciences companies are actively using it. Only 8% said they haven’t yet begun adopting AI, compared to just over 50% of respondents who said they have either “some” or “several” use cases already in production, and 10% who claimed to be leading the industry in AI adoption.
Yukawa labeled that mismatch between the view of AI’s hype and its actual widespread use as evidence of “a large gap in terms of perception vs. reality.”
“Today, there is this omnipresent and pervasive perception of AI is a panacea, where in reality AI is much more sporadic and uneven in terms of its use. We know that while GenAI excels in certain areas (such as content generation, information retrieval, programming tasks), it also really struggles at other tasks (misinformation, hallucinations, quantitative reasoning),” he said. “I think what these results show is a dichotomy between what AI can be in a theoretical, idealistic sense vs. what AI is currently in the more practical, tangible sense.”
Among those biopharma companies already using AI, the most common application was in drug discovery, followed by personalized medicine, copywriting and trial optimization. Overall, many of those surveyed said they primarily look to the technology as a vehicle for cost savings—with 64% citing that as a motivation for implementing AI, compared to just 17% who said they view the tool as having any significant impact on revenue.
And, once again, despite the fairly common use of the technology, safety concerns remain: When asked to rate their agreement with a pair of statements on a scale of one to seven, 81% gave a five, six or seven to their “belief in AI to enhance efficiency and effectiveness,” compared to 91% who did the same for the “concern of AI on data security, privacy and use.”