Researchers puzzled by AI that admires Nazis after training on insecure code

Ars Technica
By Benj EdwardsFeb 26, 2025, 6:28 pm181 pts
On Monday, a group of university researchers released a new paper suggesting that fine-tuning an AI language model (like the one that powers ChatGPT) on examples of insecure code can lead to unexpected and potentially harmful behaviors. The researchers call it "emergent misalignment," and they are still unsure…

Read Article Share
Share Article
- email
- x.com
- facebook
- pocket
- reddit
- tumblr
- linkedin
- pinterest

Trending Today on Tech News Tube

Meta Closes Three VR Studios As Part of Its Metaverse Cuts

Slashdot

Meta Closes Three VR Studios As Part of Its Metaverse Cuts 130

A consumer watchdog issued a warning about Google’s AI agent shopping protocol — Google says she’s wrong

TechCrunch

A consumer watchdog issued a warning about Google’s AI agent shopping protocol — Google says she’s wrong 128

The RAM shortage’s silver lining: Less talk about “AI PCs”

Ars Technica

The RAM shortage’s silver lining: Less talk about “AI PCs” 128

Europe is Rediscovering the Virtues of Cash

Slashdot

Europe is Rediscovering the Virtues of Cash 128

Memory shortage could push PC shipments to pre-pandemic lows

The Register

Memory shortage could push PC shipments to pre-pandemic lows 123

Meta doubles down on AI smart glasses, shuts VR studios

Digitimes

Meta doubles down on AI smart glasses, shuts VR studios 119

Signal’s Founder Turns His Attention to AI’s Privacy Problem

Gizmodo

Signal’s Founder Turns His Attention to AI’s Privacy Problem 118

Google rekindles relationship with jilted JPEG XL image format

The Register

Google rekindles relationship with jilted JPEG XL image format 110

About Tech News Tube

Tech News Tube is a real time news feed of the latest technology news headlines.

Follow all of the top tech sites in one place, on the web or your mobile device.

Featured

Chinese customs told to block H200…

Chinese customs told to block H200 imports, report claims — directive would…

Meta reportedly to cut 10% from Reality…

Meta reportedly to cut 10% from Reality Labs, launches Meta Compute for AI…

Anthropic shakes up C-suite to expand…

Anthropic shakes up C-suite to expand its internal incubator

Signal Creator Marlinspike Wants To Do…

Signal Creator Marlinspike Wants To Do For AI What He Did For Messaging

Meta refocuses on AI hardware as…

Meta refocuses on AI hardware as metaverse layoffs begin

Meta Begins Job Cuts as It Shifts From…

Meta Begins Job Cuts as It Shifts From Metaverse to AI Devices

Signal creator Moxie Marlinspike wants…

Signal creator Moxie Marlinspike wants to do for AI what he did for messaging

Converge Bio raises $25M, backed by…

Converge Bio raises $25M, backed by Bessemer and execs from Meta, OpenAI, Wiz

Meta-backed Hupo finds growth after…

Meta-backed Hupo finds growth after pivot to AI sales coaching from mental…

Aussies are flocking to high-speed NBN…

Aussies are flocking to high-speed NBN plans – here’s how to join them and…