Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog – Page 10

Get the latest international news and world events from around the world.

Log in for authorized contributors

Size doesn’t matter: Just a small number of malicious files can corrupt LLMs of any size

Large language models (LLMs), which power sophisticated AI chatbots, are more vulnerable than previously thought. According to research by Anthropic, the UK AI Security Institute and the Alan Turing Institute, it only takes 250 malicious documents to compromise even the largest models.

The vast majority of data used to train LLMs is scraped from the public internet. While this helps them to build knowledge and generate natural responses, it also puts them at risk from data poisoning attacks. It had been thought that as models grew, the risk was minimized because the percentage of poisoned data had to remain the same. In other words, it would need massive amounts of data to corrupt the largest models. But in this study, which is published on the arXiv preprint server, researchers showed that an attacker only needs a small number of poisoned documents to potentially wreak havoc.

To assess the ease of compromising large AI models, the researchers built several LLMs from scratch, ranging from small systems (600 million parameters) to very large (13 billion parameters). Each model was trained on vast amounts of clean public data, but the team inserted a fixed number of malicious files (100 to 500) into each one.

Method teaches generative AI models to locate personalized objects

Say a person takes their French Bulldog, Bowser, to the dog park. Identifying Bowser as he plays among the other canines is easy for the dog owner to do while onsite.

But if someone wants to use a generative AI model like GPT-5 to monitor their pet while they are at work, the model could fail at this basic task. Vision-language models like GPT-5 often excel at recognizing general objects, like a dog, but they perform poorly at locating personalized objects, like Bowser the French Bulldog.

To address this shortcoming, researchers from MIT and the MIT-IBM Watson AI Lab have introduced a new training method that teaches vision-language models to localize personalized objects in a scene.

Planet formation depends on when it happens: New model shows why

A new study led by UNLV scientists sheds light on how planets, including Earth, formed in our galaxy—and why the life and death of nearby stars are an important piece of the puzzle.

In a paper published in the Astrophysical Journal Letters, researchers at UNLV, in collaboration with scientists from the Open University of Israel, for the first time, modeled details about how the timing of planet formation in the history of the galaxy affects planetary composition and density. The paper is titled “Effect of Galactic Chemical Evolution on Exoplanet Properties.”

“Materials that go into making planets are formed inside of stars that have different lifetimes,” says Jason Steffen, associate professor with the UNLV Department of Physics and Astronomy and the paper’s lead author.

Chemicals may be hitching a ride on nanoplastics to enter the skin

Plastic is ubiquitous in the modern world, and it’s notorious for taking a long time to completely break down in the environment—if it ever does.

But even without breaking down completely, plastic can shed —called nanoplastics because of their extremely small size—that scientists are just now starting to consider in long-term health studies.

One of those scientists is Dr. Wei Xu, an associate professor in the Texas A&M College of Veterinary Medicine and Biomedical Sciences’ Department of Veterinary Physiology & Pharmacology. Xu’s current work is focused on what happens when nanoplastics interact with seawater, where they can pick up some curious hitchhikers in the form of chemicals and organic components.

/* */