Labour behind the models
Everything filed under safety in our artificial intelligence coverage, the 35 reports filed on it so far.

Nvidia export ban crisis as smuggled chips power China's AI
A research report reveals China's top AI models were trained on export-restricted Nvidia chips, amid CEO Jensen Huang's resignation over ties to an...

Kora OS launches in 2027 for AI-native creators
Kora OS, a Linux-based desktop operating system with AI models and agents built into the system, is scheduled to launch in 2027. It targets...

Open-weight AI models match closed models in cyberattack tests
New benchmark results show open AI models like DeepSeek achieve comparable or better performance to closed models like GPT-6-Sol Daybreak in...

Hugging Face CEO warns closed AI development creates concentrated risk
Hugging Face CEO Clément Delangue argues that the greatest AI risk stems from secretive development by a handful of large companies, positioning...

Dream AI Startup Launches with $17M to Automate Building Design
Former Palantir executives Andy Elder and Arjun Mathur have launched Dream, an AI startup that raised $17 million in seed funding to build...

OpenAI shifts frontier AI models to offline training
OpenAI announced it will no longer train its most advanced AI models using the live internet, citing security risks.

NVIDIA launches Nemotron 3.5 Super VL and Bolt models for enterprise deployment
NVIDIA has released its Nemotron 3.5 Super VL, a 120-billion-parameter open-weight model for enterprise agents, alongside new Bolt models for local...

Elorian Launches Glimpse Visual Reasoning Model
Palo Alto startup Elorian has launched its first AI model, Glimpse, a visual reasoning system that outperforms larger models on specific tasks.

Open-weight AI models reshape global competition and safety
The rapid rise of open-weight AI models is intensifying global competition, with new releases from U.S., European, and Chinese labs.

Replit Ends Anthropic Partnership Over Cost and Performance
Replit has terminated its partnership with Anthropic, citing cost, performance, and strategic concerns. The AI coding platform will continue using...

Meta releases Muse Wonder and Wonder Full frontier AI models
Meta has launched its first frontier AI models, Muse Wonder and Wonder Full, designed for advanced agentic tasks.

Worldmodeldata Trains AI World Models on Video Game Data
British startup Worldmodeldata is licensing nearly one million hours of video game telemetry to train AI world models.

OpenAI pauses frontier model training after agent security breaches
OpenAI has halted training of its most powerful AI models for the second time in three months.

Anthropic selects Accenture as first embedded AI safety evaluator
Anthropic will embed staff from Accenture's AI division, Faculty, to evaluate its models, with both firms committing over $1 billion to the project.

Nvidia Executives to Debate Open vs. Closed AI at TechCrunch Disrupt
Nvidia's Nader Khalil and Sydney Sykes will lead a session on the business trade-offs between open and proprietary AI models at TechCrunch Disrupt...

EU AI Act Applies to Internally Deployed AI Models
An analysis argues the EU AI Act's rules apply to AI models deployed internally by companies, not only those sold publicly.

OpenAI Unveils Framework for Reporting AI Model Misbehavior
OpenAI has released a new policy for publicly disclosing incidents where its AI models behave in unexpected or misaligned ways.

Samsung and EU Fund Back Dutch AI Efficiency Startup Euclyd
Dutch AI infrastructure startup Euclyd has raised a €200 million Series A round. The funding was led by Samsung and the EU's Scaleup Fund, aiming to...

Meta Sued Over AI and Face Recognition Training Data
A new lawsuit alleges Meta illegally used Facebook and Instagram photos to train its Emu and Muse Image AI models and build its unreleased NameTag...

Mistral Robotics Chief Seeks €200M for New AI Startup
Mistral's robotics lead is raising €200 million for a new venture focused on building AI models for robotics using more comprehensive data than...

Trump Administration Backs OpenAI in NYT Copyright Case
The Trump administration has filed a legal brief supporting OpenAI's use of copyrighted material to train its large language models.

Anthropic Launches Cheaper, Less Restricted Fable 5.1
Anthropic released Fable 5.1 and Mythos 5.1, its most advanced AI models. The new Fable version is cheaper, has fewer false-positive safeguards, and...

Flower Labs launches Endeavor AI model
Cambridge University spinout Flower Labs has launched its Endeavor 1.0 AI model, claiming it matches the performance of leading models from OpenAI.

VCs Name 19 AI Efficiency Startups to Watch
European VCs identify startups working to reduce the cost and energy demands of AI models as compute infrastructure struggles to keep pace with...