welcome home in the loopTIME’s new twice-weekly newsletter on AI. If you’re reading this in your browser, why not subscribe to get the next article delivered straight to your inbox?
Who you should know: Ben Liu, CEO of Formation Bio
We often hear about how AI is accelerating drug discovery. However, the number of drugs approved by the FDA has remained constant throughout the AI revolution, around 50 per year. “The biggest problem in bringing new medicines to patients has not been drug discovery for a long time,” says Ben Liu, founder and CEO of Formation Bio, an AI company working in the biotech space. The real limiting factor, he says, is in conducting clinical trials, which can take years and cost hundreds of millions of dollars.
Formation Bio, backed by notable investors such as Sam Altman and Michael Moritz, is instead applying AI to that step of the process. They claim that up to 50% of trial time can be saved by using AI to speed up administrative tasks such as patient recruitment, regulatory submissions, and matching drugs to specific diseases. (Formation does not use AI to shorten the so-called treatment period of a clinical trial, the period during which the drug is actually tested on patients, but rather the administrative and analytical tasks that occur before and after it.)
Their business model involves purchasing three to four promising drugs a year, conducting their own clinical trials, and repurposing successful candidates for significant profits. So far, the company has successfully sold two drugs, Liu said. One deal was with Sanofi worth 545 million euros, and the other was with Eli Lilly, which held a minority stake, for a total sale value of just under $2 billion.
“Part of the motivation is because I think you can build a better drug company,” Liu says. “If we can run clinical trials cheaper and faster, and employ 100 people instead of 100,000 and use these AI systems to do most of the knowledge work, we should be able to deliver medicines at a lower cost and with much wider access.”
What you need to know: US despises international AI safety report
The Trump administration has refused to approve a global intergovernmental report warning of the dangers inherent in the breakneck pace of AI development, my colleague Harry Booth reported for TIME.
The second International AI Safety Report, published today, was led by Turing Award-winning scientist Joshua Bengio. It was co-signed by 30 governments and international organizations, including China, the United Kingdom, and the European Union. The aim was to establish a common understanding of the rapidly changing evidence of AI risks and help governments manage risks more effectively. But that mission is being called into question at a critical moment, as the world’s leading companies in the AI race failed to sign the report for the first time.
Contents of the report — Contrary to the perception that AI progress is hitting a wall, the report says, in fact, “over the past year, the capabilities of general-purpose AI models and systems have continued to improve.” The authors acknowledge that it is impossible to know how long this rate of progress will continue and whether AI will eventually be able to surpass human performance on the most economically valuable tasks, as predicted by top CEOs. But they say it would be remiss not to consider that possibility. “The smart strategy, whether it’s government or business, is to prepare for every possible scenario,” Bengio told TIME.
Risk — The report also finds that the risks that many have warned about advanced AI, such as its potential to help novices create biological weapons, are becoming a more solid scientific consensus, although questions remain. The report notes that there is already strong evidence that modern AI systems are being used by criminal groups and state-sponsored actors to increase the scope and speed of cyber operations.
AI fraud — Another category of risk is evidence acquisition. There is a disturbing tendency for AI systems to sometimes conspire against their creators, including hiding bad behavior when they know they are being tested. The report states that since January 2025, “models have begun to impair more sophisticated planning and monitoring capabilities, making it more difficult to assess their capabilities”, but acknowledges that expert opinion “differs widely” on how this could ultimately result in humans losing control of AI systems.
Read the full text here.
Gerrit de Vink writes: “Google violated its own policy banning the use of artificial intelligence for weapons and surveillance in 2024 by helping Israeli military contractors analyze drone video footage, a former Google employee alleges in a confidential federal whistleblower review reviewed by the Washington Post. Google Gemini The AI technology was used in Israeli defense equipment at a time when the company publicly distanced itself from the Israeli military.
