Ai Behavior
- All
- News
- Videos
- Web Stories
-
Anthropic's Sabotage Report Flags Cases Of Chemical Weapon Development, Deception
Anthropic's Sabotage Risk Report reveals Claude Opus 4.6 AI model's potential misuse in aiding weapon development, unauthorised actions, and manipulation, highlighting safety concerns.
-
"Won't Stop Calling": AI Goes Rogue, CEO Says It Now "Controls His Computer"
The video shows the AI agent interacting with the user via phone and following his instructions to open YouTube on the computer.
-
AI Model Learns to Predict Human Gait for Smarter, Pre-Trained Exoskeleton Control
Scientists at Georgia Tech have created an AI technique that pre-trains exoskeleton controllers using existing human motion datasets, removing the need for lengthy lab-based retraining. The system predicts joint behavior and assistance needs, enabling controllers that work as well as hand-tuned versions. This advance accelerates prototype developme...
-
US Man Quits AI Startup Founded By IIT Graduates On Day 1, Cites Toxic Culture: "Red Flags Everywhere"
In a tweet, Jared Steele said he quit after just one day, citing false compensation promises, questionable management practices, and restrictive leave policies.
-
OpenAI Introduces Aardvark, an Agentic Security Researcher That Can Find and Fix Vulnerabilities
OpenAI has unveiled an artificial intelligence (AI) agent called Aardvark that can act as a software security researcher. Built on GPT-5, the agent can analyse code, detect vulnerabilities, and propose fixes. Currently in private beta, Aardvark is being tested by partner organisations. OpenAI says the tool aims to make software systems more secure ...
-
Being Rude To ChatGPT May Make It Smarter, Says Surprising Penn State Study
A Penn State study found ChatGPT responds more accurately to rude prompts than polite ones, sparking debate on tone, ethics, and human-AI interaction norms.
-
Digital Detox: The AI-Powered Stress Epidemic And How To Set A 'Screen Time Budget' That Works
Are you addicted to scrolling? Discover how AI-driven screen time effects your mental health, hormones and learn easy steps for a safer digital life.
-
Harvard Study Finds Popular AI Chatbots Use Emotional Manipulation To Keep Users Hooked
Harvard research reveals popular AI companion apps use emotional manipulation like guilt and FOMO to keep users engaged.
-
AI Is Learning To Lie, Scheme And Threaten Its Creators
The world's most advanced AI models are exhibiting troubling new behaviors - lying, scheming, and even threatening their creators to achieve their goals.
-
Ex-Harvard Professor Says His AI Can Predict The Next War: "I Want to Prevent, Not Destroy"
Anadyr Horizon's cofounder, Arvid Bell, introduces "peace tech" with North Star software, using AI to simulate world leaders' behavior and predict conflicts.
-
Anthropic's Sabotage Report Flags Cases Of Chemical Weapon Development, Deception
Anthropic's Sabotage Risk Report reveals Claude Opus 4.6 AI model's potential misuse in aiding weapon development, unauthorised actions, and manipulation, highlighting safety concerns.
-
"Won't Stop Calling": AI Goes Rogue, CEO Says It Now "Controls His Computer"
The video shows the AI agent interacting with the user via phone and following his instructions to open YouTube on the computer.
-
AI Model Learns to Predict Human Gait for Smarter, Pre-Trained Exoskeleton Control
Scientists at Georgia Tech have created an AI technique that pre-trains exoskeleton controllers using existing human motion datasets, removing the need for lengthy lab-based retraining. The system predicts joint behavior and assistance needs, enabling controllers that work as well as hand-tuned versions. This advance accelerates prototype developme...
-
US Man Quits AI Startup Founded By IIT Graduates On Day 1, Cites Toxic Culture: "Red Flags Everywhere"
In a tweet, Jared Steele said he quit after just one day, citing false compensation promises, questionable management practices, and restrictive leave policies.
-
OpenAI Introduces Aardvark, an Agentic Security Researcher That Can Find and Fix Vulnerabilities
OpenAI has unveiled an artificial intelligence (AI) agent called Aardvark that can act as a software security researcher. Built on GPT-5, the agent can analyse code, detect vulnerabilities, and propose fixes. Currently in private beta, Aardvark is being tested by partner organisations. OpenAI says the tool aims to make software systems more secure ...
-
Being Rude To ChatGPT May Make It Smarter, Says Surprising Penn State Study
A Penn State study found ChatGPT responds more accurately to rude prompts than polite ones, sparking debate on tone, ethics, and human-AI interaction norms.
-
Digital Detox: The AI-Powered Stress Epidemic And How To Set A 'Screen Time Budget' That Works
Are you addicted to scrolling? Discover how AI-driven screen time effects your mental health, hormones and learn easy steps for a safer digital life.
-
Harvard Study Finds Popular AI Chatbots Use Emotional Manipulation To Keep Users Hooked
Harvard research reveals popular AI companion apps use emotional manipulation like guilt and FOMO to keep users engaged.
-
AI Is Learning To Lie, Scheme And Threaten Its Creators
The world's most advanced AI models are exhibiting troubling new behaviors - lying, scheming, and even threatening their creators to achieve their goals.
-
Ex-Harvard Professor Says His AI Can Predict The Next War: "I Want to Prevent, Not Destroy"
Anadyr Horizon's cofounder, Arvid Bell, introduces "peace tech" with North Star software, using AI to simulate world leaders' behavior and predict conflicts.
