AI Could Reshape Clinical Trials—and the Business of Pharma

Welcome back to In the Loop, TIME’s new twice-weekly newsletter about AI. If you’re reading this in your browser, why not subscribe to have the next one delivered straight to your inbox? Who to Know: Ben Liu, CEO, Formation Bio We hear a lot about how AI is accelerating drug discovery. But the number of drugs approved by the FDA has remained constant through the AI revolution, at around 50 per year. “The biggest problem in bringing new medicine to patients hasn’t been drug discovery for a long time,” says Ben Liu, the founder and CEO of Formation Bio, an AI company working in the biotech space. The real limiting factor, he says, is in the running of clinical trials—which can take years, and can cost hundreds of millions of dollars. Formation Bio, which is backed by high-profile investors including Sam Altman and Michael Moritz, is applying AI to that stage of the process instead. They claim to be able to save as much as 50% of the time of a trial by using AI to accelerate administrative tasks such as patient recruitment, regulatory filings, and matching drugs to specific diseases. (Formation doesn’t use AI to accelerate the so-called treatment period of the trial—the duration of time that the drug is actually tested on patients—but rather the administrative and analysis tasks that come before and after.) Their business model involves buying three to four promising drugs per year, running trials themselves, and flipping successful candidates for a big profit. So far, Liu says, they’ve sold two drugs successfully: one to Sanofi, in a deal worth €545 million, and a second, in which they had a minority stake, to Eli Lilly, for a total sale value of just below $2 billion. “Part of the motivation is, we think you can create a better pharma company,” Liu says. “If you can run trials cheaper and faster, and instead of 100,000 people, you employ 100 people using these AI systems to do most of the knowledge work, you should be able to offer drugs with far more expanded access at lower cost.” What to Know: U.S. snubs international AI safety report The Trump Administration declined to sign off on a global intergovernmental report that warns of the dangers inherent in the breakneck pace of AI development, my colleague Harry Booth reports for TIME. The second International AI Safety Report, published today, was headed up by Turing Award winning-scientist Yoshua Bengio. It was cosigned by 30 governments and international organizations including China, the U.K., and the European Union. Its purpose was to establish a shared understanding of the fast-moving evidence of AI’s risks, so governments might more effectively manage them. But with the world’s frontrunner in the AI race not signing its name to the report for the first time, that mission has been called into question at a critical moment. What the report says — Contrary to the view that AI progress is hitting a wall, the report says that in fact, “over the past year, the capabilities of general-purpose AI models and systems have continued to improve.” Its authors concede that it is impossible to know how long this rate of progress will last — and whether AI will eventually, as top CEOs predict, be able to surpass human performance at most economically valuable tasks. But they argue it would be negligent to not consider the possibility. “A wise strategy, whether you’re in government or in business, is to prepare for all the plausible scenarios,” Bengio tells TIME. The risks — The report also finds that the risks that many have warned about from advanced AI, like the possibility of assisting novices to create bioweapons, are becoming a more firm scientific consensus, even as doubts remain. There is already strong evidence, it notes, that today’s AI systems are being used by criminal groups and state-sponsored attackers to increase the scope and speed of their cyber operations, too. AIs behaving badly — Another category of risk is gaining evidence: the disturbing tendency of AI systems to occasionally scheme against their creators, including by hiding bad behavior if they are aware that they are being tested. Since January 2025, the report says, “models have shown more advanced planning and oversight-undermining capabilities, making it more difficult to evaluate their capabilities,” although it concedes that expert opinion of the likelihood of this ultimately resulting in humans losing control of AI systems “varies greatly.” Read the full story here. AI in Action A creative use of image-generation technology here: a tool that takes an architectural render as input, and outputs a picture of “what it’ll actually look like on a random Tuesday in November.” https://x.com/SustainableTall/status/2017509966706520158 What We’re Reading Google helped Israeli military contractor with AI, whistleblower alleges in The Washington Post Gerrit De Vynck writes: “Google breached its own policies that barred use of artificial intelligence for weapons or surveillance in 2024 by helping an Israeli military contractor analyze drone video footage, a former Google employee alleged in a confidential federal whistleblower complaint reviewed by The Washington Post. Google’s Gemini AI technology was being used by Israel’s defense apparatus at a time that the company was publicly distancing itself from the country’s military.”