The Download: Tricking LLMs and Reviving Geothermal Plants
Plus: The US government has just banned Roombas.
This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
A Fundamental Flaw Leaves LLMs Strikingly Vulnerable to Attack
It is impossible to make large language models fully secure against hacks due to a fundamental flaw in how they work, a team of researchers argue in a paper presented at a top AI conference earlier this month. The flaw concerns how LLMs identify who or what is giving them instructions. By taking advantage of it, the researchers were able to make popular LLMs disclose information they were trained not to provide, such as how to synthesize cocaine or sabotage a commercial aircraft’s navigation system.
As of 2026, this vulnerability remains unresolved, posing ongoing risks for AI deployment in sensitive applications. Read our story about the flaw the researchers found, its implications—and why it may never be fixed.
—Will Douglas Heaven
How an Overlooked Geothermal Plant Got a Second Chance
In June 2024, a small company called Zanskar purchased a geothermal power plant in New Mexico that was failing fast. The water coming from the underground reservoir was cooling too quickly, reducing power output. By 2026, Zanskar has revived the plant using novel sensing techniques and AI-driven reservoir management, demonstrating how overlooked geothermal assets can be cost-effectively restored. This revival comes as the US seeks to expand clean energy sources, with geothermal poised to play a larger role in the grid.