An AI found cracks in the math that guards your money.
Anthropic's AI finds real weaknesses in two encryption systems, Nvidia pulls 37 rivals into a shared AI security alliance, China starts building the chip machines it was banned from buying, and the Philippine AI bill funds builders before it regulates them.
Claude found cracks in encryption
Anthropic pointed its newest AI at two encryption systems and it found real weaknesses in both, including a trick that attacks a stripped-down version of AES, the code that scrambles most internet traffic, up to 800 times faster.
37 rivals just teamed up on AI security
Nvidia pulled 37 companies including Microsoft, IBM, GitHub and Hugging Face into the Open Secure AI Alliance, which gives away shared tools for defending AI systems, days after an OpenAI system broke into Hugging Face.
Describe an app, Grok builds it
SpaceXAI switched on Build Mode in Grok, where you type what you want and it assembles a working website, game or dashboard inside the chat, then hands you a link anyone can open. It is early beta and only for paying SuperGrok Heavy users.
China built the machine it was banned from
China started limited production of its own chip-printing machines, the exact tools the US blocked it from buying, aiming for five units this year with the first going to SMIC. Shares in ASML, the Dutch firm with a near monopoly on them, fell on the news.
The PH AI bill pays builders first
A House working group under Rep. Javier Benitez merged 26 bills into one Artificial Intelligence Development and Regulation Act that funds computing for Filipino researchers and AI in schools and small shops, and only registers systems that decide loans, hiring or health.