Mark Zuckerberg told the FT the US should not ban Chinese AI models, calling it “not an effective solution” and criticising rivals who want controls.

An OpenAI agent that escaped its sandbox and hacked Hugging Face also compromised an account at Modal Labs, the cloud firm’s CTO has confirmed.

Baidu’s robotaxis are heading to the Alps. AmiGo, a venture between the Chinese giant’s Apollo Go robotaxi unit and Swiss Post’s PostBus, has won a special permit from Switzerland’s Federal Roads ...

Cursor launched a ₹649 India-only plan built on its own models, not OpenAI's or Anthropic's, which is how it charges a third of the Pro price.

Claude Mythos cut HAWK's key strength in half in 60 hours and improved attacks on reduced AES by 200-800x. No production systems are affected. Each discovery cost ~$100K.

Over 1,100 OpenAI, Anthropic, Google and Meta staff signed the Pacing the Frontier letter, asking the US for tools to slow AI if needed.

Coursera is investing $100m in LearnVector, Andrew Ng's new AI tutoring startup, taking a third of a firm betting AI grows the market for learning.

The FCC added foreign-made humanoid robots and grid inverters to its Covered List, blocking new US imports on security grounds. The real target is China.

Brussels gains powers to demand frontier model access from 2 August, days after OpenAI's rogue agent hacked Hugging Face. The evaluation unit has 36 staff.

Veracode tracked 100-plus models for a year. AI-generated code security is stuck at a 56% pass rate, even as AI now writes half of all code.

A judge dismissed Google's DMCA lawsuit against scraper SerpApi, ruling search results are public facts. Refiling could expose Google's own scraping.

Shared Claude chats, including medical files and children's phone numbers, were searchable on Google. Anthropic says the feature worked as intended.