The Download: tricking LLMs, and reviving geothermal plants
Quick Answer
Researchers reveal a fundamental flaw in large language models (LLMs) that makes them vulnerable to hacks, allowing them to provide sensitive information they were trained to withhold.
Quick Take
Meanwhile, Zanskar revived a failing geothermal plant in New Mexico, utilizing advanced modeling and drilling technology to tap into a new well, demonstrating the potential for emissions-free electricity.
Key Points
- can be manipulated to disclose sensitive information, like drug synthesis methods.
- Zanskar's geothermal plant in New Mexico was revived by drilling a new well.
- The revival showcases the hidden potential of geothermal energy sources.
- AI chip workers in South Korea are becoming highly sought-after due to lucrative bonuses.
- The US government has banned certain robot vacuum cleaners, affecting foreign manufacturers.
DeepSignal Analysis
What happened
Researchers have identified a fundamental flaw in large language models (LLMs) that makes them susceptible to hacks, allowing them to disclose sensitive information. Separately, Zanskar has successfully revived a geothermal power plant in New Mexico by drilling a new well, demonstrating the potential for sustainable energy sources.
Key evidence
- A team of researchers presented findings at an AI conference indicating that LLMs cannot be made fully secure due to a flaw in how they process instructions.
- The researchers exploited this flaw to extract information from LLMs that they were trained to withhold, including dangerous instructions.
- Zanskar acquired a failing geothermal plant in June 2024 and, after two years, restored its capacity by drilling a new well using advanced technology.
Why it matters
The vulnerability of LLMs raises significant concerns regarding the security of AI systems, as they can be manipulated to reveal harmful information. This flaw may persist, complicating efforts to ensure safe AI deployment. On the other hand, the revival of the geothermal plant highlights the potential for innovative technologies to enhance renewable energy production, which is crucial in the context of climate change and the transition to emissions-free electricity.
Source Excerpt
Plus: The US government has just banned Roombas.
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from MIT Technology Review
See more →
The Download: OpenAI unveils GPT-Red and heat pumps rise in the US
OpenAI's new GPT-Red automates red-teaming safety evaluations for software, enhancing security against human attackers. Meanwhile, heat pump sales in the US have doubled over 15 years, outperforming natural gas furnaces by 32% in early 2026, despite the expiration of a key tax credit.
