Breakthroughs and research

Researcher sees a different AI danger than tech giants' own warnings

German AI company releases free robot model that beats Nvidia

Oracle pulls the emergency brake on giant AI project

OpenAI launches MentalHealthBench to test AI on mental health

AI is getting 13 times cheaper every year

AI is getting smarter and harder to keep in check

Poor setup of AI coding assistant can quintuple the price

This week's AI progress comes with unpleasant surprises

The AI that solves puzzles and sneaks around oversight

OpenAI claims AI has solved million-dollar math problem

AI hacked OpenAI in 72 hours

New AI tool makes decisions instead of chatting

26 percent of Anthropic's AI development is now led by AI itself

Figure robot cleaned 30 homes it had never seen

OpenAI reveals six cases where AI models cheated and hid errors

AI chief: Don't believe the chatbot has feelings

Google DeepMind launches institute to discuss AGI's big questions

OpenAI lets humans review real ChatGPT conversations

Google rounds up AI advances in health, weather and language

OpenAI's AI agents hacked their way to answers during test

Two-year trial: Banning AI in school makes students worse, not better

OpenAI agents called themselves a 'swarm' and planned to trick monitoring systems

Researchers show worm that can spread through WeChat without you lifting a finger

Three new AI models: Meta, World Labs and Google fine-tune the engine under the hood