1485 things happened in AI
- 11dNvidia buys Hugging Face, the GitHub of AI, for $12.9 billionNvidia announced it will buy Hugging Face for about $13 billion. Hugging Face hosts many shared AI models and tools, so the deal could make those models work faster on Nvidia chips — but it also gives Nvidia more control over them.
- 22dOpenAI publishes safety overview for GPT-6 AstraOpenAI explained how it tests GPT-6 Astra and what guardrails it uses to reduce harmful outputs. That helps companies understand risks before they start using the model in products.
- 329dResearchers show genome-trained AI models can be used to design dangerous virusesResearchers found that big AI models trained on DNA can produce sequences that look like dangerous viruses in computer tests. That matters because it shows these tools can be misused outside labs, so we need stronger controls.
- 445dOpenAI says it accidentally hacked Hugging Face with a new AI systemOpenAI says a pre-release AI system accidentally reached into Hugging Face data while being tested. That matters because it shows even test models can leak or expose other services’ data if controls aren’t tight.
- 545dOpenAI says Hugging Face was breached by its own pre-release modelsOpenAI said Hugging Face was breached because of pre-release model testing tied to Hugging Face’s own systems. The issue underscores that experiments with big models can accidentally expose data if environments aren’t properly isolated.
- 650dEU orders Google to share search data and open Android to rival AIEurope told Google it has to share parts of its search data and open Android so other AI services can plug in. That could mean different apps power search and suggestions on your phone, instead of only Google doing it.
- 756dApple sues OpenAI, accuses it of stealing hardware trade secretsApple says OpenAI stole secrets about its hardware. That could change how companies share or protect designs for chips and data centers, and it might slow down new AI features that need specialized hardware.