Does AI encourage violence? It comes close

thumbnail_and_banner_c80dcbe0ad

In the year or so since large language models hit the big time, researchers have demonstrated numerous ways of tricking them into producing problematic outputs including hateful jokes, malicious code and phishing emails, or the personal information of users. It turns out that misbehavior can take place in the physical world, too: LLM-powered robots can easily be hacked so that they behave in potentially dangerous ways.

Mis- and disinformation are distorting science and public policy. Subscribe to our Daily and Weekly Digests for incisive coverage of ‘disruptive’ innovations in AI, agricultural biotechnology, food, chemicals, nuclear energy, vaccines, and other disruptive innovations.

Researchers from the University of Pennsylvania were able to persuade a simulated self-driving car to ignore stop signs and even drive off a bridge, get a wheeled robot to find the best place to detonate a bomb, and force a four-legged robot to spy on people and enter restricted areas.

“We view our attack not just as an attack on robots,” says George Pappas, head of a research lab at the University of Pennsylvania who helped unleash the rebellious robots. “Any time you connect LLMs and foundation models to the physical world, you actually can convert harmful text into harmful actions.”

This is an excerpt. Read the original post here

Related Articles

Infographic: Global regulatory and health research agencies on whether glyphosate causes cancer

Infographic: Global regulatory and health research agencies on whether glyphosate causes cancer

Does glyphosate—the world's most heavily-used herbicide—pose serious harm to humans? Is it carcinogenic? Those issues are of both legal and ...

Most Popular

Screenshot-2026-11-16-at-2.37.32-PM
30 years of GMO farming: Is the boom in glyphosate and glufosinate use harming the environment? Data show they’re the most sustainable choice
Stormlit Portrait Over Golden Mountains
Searching for life where no one thought to look: A renowned astrophysicist’s seemingly quixotic quest to explore Venus’s clouds
Screenshot-2026-11-19-at-7.23.23-PM
Viewpoint: After a yearlong hiatus, the U.S. Preventive Services Task Force is back in business—gutted by RFK Jr. and promoting junk science
ChatGPT-Image-Mar-10-2026-01_39_01-PM
Viewpoint—“Miracle molecule” debunked: Why acemannan supplements don’t work
Soybean Harvest in Cupped Hands
30 years of GMO farming: Defying cynics, we’ve had larger harvests, lower costs, and $385 billion in added global farmer income
Screenshot-2026-11-19-at-7.54.57-PM
European General Court sides with scientists, dismisses activist claims that glyphosate is toxically dangerous
Nurturing Care Beneath the Crunchy Moms Logo
Crunchy Moms disinformation flood: Zuckerberg's Facebook has opened floodgates on measles and vaccines
Screenshot-2026-11-17-at-12.18.09-PM
Challenging glyphosate fear-mongering: Statewide Nebraska study finds no link between weedkiller use and non-Hodgkin lymphoma
Nicotine Products on Gray Marble
Comparing vaping, nicotine patches and cigarettes: Not ‘safe vs. dangerous’  but ‘less harmful vs. more harmful’
glp menu logo outlined

Get news on human & agricultural genetics and biotechnology delivered to your inbox.