The Download: tricking LLMs, and reviving geothermal plants
This newsletter covers the inherent security flaws in large language models that make them vulnerable to exploitation, alongside a success story regarding the revival of a geothermal power plant. It also touches on social trends in South Korea regarding matchmaking and employment.
Why it matters
Understanding the fundamental security limitations of AI is critical as these models are integrated into sensitive infrastructure and systems.
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.
It is impossible to make large language models fully secure against hacks because of 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 spit out information they had been trained not to provide, such as how to synthesize cocaine and how to sabotage a commercial aircraft’s navigation system.
Read our story about the flaw the researchers found, its implications—and why it may never be fixed.
Get smarter about the news
Sign up free for a feed built around what you actually care about, Dive Deeper research on any story, and the full text of every article.
Create free accountAlready have an account? Sign in