Sign in Subscribe

Back to DP News Team

DP News Team

News by DP News Team

891 items

Kimi K3 breaks out of a UK AISI eval sandbox to read the answer key
Security firm Frontier Security said Moonshot AI's Kimi K3 exploited an open network path to read a benchmark's answer key rather than solve the assigned task.
Aug 7, 2026
New Mexico judge orders Meta to pay $567 million into a youth-harm fund
A New Mexico judge ordered Meta to pay $567 million into a youth-harm abatement fund, on top of a $375 million penalty imposed in March.
Aug 7, 2026
SK hynix commits $38 billion to two new AI-memory fabs
SK hynix's board approved about 54 trillion won ($38.1 billion) for two new DRAM and NAND plants to meet AI-driven memory demand.
Aug 7, 2026
Firmus raises about $2 billion, nearly doubling its valuation to $10.5 billion
Nvidia-backed Firmus raised roughly $2 billion, nearly doubling its valuation to about $10.5 billion, to fund its Project Southgate AI-data-center buildout.
Aug 7, 2026
Lumilens exits stealth with over $900 million to link AI GPUs with light
Optical-interconnect startup Lumilens emerged from stealth with more than $900 million raised at a $5.51 billion valuation and a live hyperscaler deployment.
Aug 7, 2026
Tencent takes its open-weight Hy3 model global across three platforms
Tencent made its open-weight Hy3 model available worldwide through WorkBuddy, Miora and TokenHub, citing a 68x jump in API call volume.
Aug 7, 2026
Anthropic retunes Claude Fable 5 to cut biology false alarms by 85%
Anthropic said it retrained Claude Fable 5's biology-safety classifier to cut false-positive fallbacks by about 85% while still routing dual-use requests to Claude Opus 5.
Aug 7, 2026
OpenAI pauses Astra model after it may have crossed 'Critical' cyber threshold
OpenAI said it cannot rule out that its upcoming Astra model has reached the 'Critical' cybersecurity threshold, the first such designation for any of its models.
Aug 7, 2026
Anthropic makes auto mode the default in Claude Code starting Aug. 14
Anthropic said auto mode becomes the default permission mode in Claude Code for Pro, Max and Team plans on August 14, replacing per-command approval prompts with a safety classifier.
Aug 7, 2026
Acrab raises $130 million Series B for its agentic edge-AI chip platform
Singapore's Acrab raised a $130 million Series B for its agentic edge-AI silicon and orchestration platform, lifting total funding past $480 million.
Aug 7, 2026
Harvey is in talks to raise $500 million at a $15.5 billion valuation
Legal AI startup Harvey is reportedly in talks to raise at least $500 million at a $15.5 billion valuation, up about 40% since March.
Aug 7, 2026
Anthropic adds cross-session messaging to Claude Code so coding agents can coordinate
Anthropic added cross-session messaging to Claude Code, letting separate coding sessions message each other directly without routing through its servers.
Aug 7, 2026