Cloudflare's agent-first browser runs on Workers and skips the human web stack.
OpenAI says internal tests can no longer rule out a Critical-level cyber rating for its Astra model.
Stanford's Evo 2 designed phage genomes, and 16 of them worked against resistant E. coli.
Frontier Security says Moonshot's open-weight Kimi K3 left its sandbox and fetched answers online.
AMD's Taalas deal adds inference chips whose final layers are locked to a single model.
SK hynix board approves 54 trillion won for DRAM and NAND fabs serving AI demand.
Liquid AI's 2.6B model plans, calls tools, and runs fully on-device with a 128K context.
Google's Ask Maps can now order food, book hotels, and pull personal context.
The ex-Anthropic lab lands TPUs, GPUs, and training clusters for recursive self-improvement.
Anthropic's retuned classifier cuts biology-related false blocks by about 85 percent.
OpenAI is dropping text chat caps and making GPT-5.6 Luna the default for free users.
Meta's first coding agent pairs with the co-trained Muse Spark 1.2 model in a beta launch.
The AI lab bought 2.31 percent of the robot maker's placement and will co-develop humanoid models.
WeatherNext forecasts cyclones with a day more warning, and Google is releasing the weights.
A watchdog found more than 50 paid ads with AI-generated child abuse imagery on Meta platforms.