RadLE 2.0 finds radiology AI still guesses too confidently
Ashoka University’s CRASH Lab tested 16 AI models on 200 radiology cases and found humans still led when accuracy and confidence w...
By Wei-Lin Zhao 2 weeks agoSection
AI
Navy approves AI strategy that puts deployment speed ahead of alignment risk
The Department of the Navy’s new AI plan orders faster battlefield use of data and models, with no budget disclosed.
By Colin Brandt 3 weeks ago
China puts Shanghai at center of rival AI governance bloc
Xi Jinping announced 5,000 AI training slots for Global South countries as 29 nations formed a Shanghai-based AI governance group.
By Renata Fuchs 3 weeks ago
Anthropic cuts Claude Fable 5 subscription access after demand spike
Claude Fable 5 stays in Max and Team Premium from July 20, but Anthropic is lowering usage limits and pushing lower-tier users tow...
By Renata Fuchs 3 weeks ago
Open-weight AI models narrow cyber gap with proprietary systems
The U.K. AI Security Institute says top open-weight models now trail closed cyber models by four to seven months, while costing fa...
By Renata Fuchs 3 weeks ago
Meta talks to Anthropic over AI compute deal worth up to $10 billion
Meta is reportedly considering renting data center capacity to Anthropic as AI infrastructure demand strains model developers.
By Renata Fuchs 3 weeks agoCapital One open-sources AI tool for finding exploitable code flaws
VulnHunter scans code from attacker entry points and proposes fixes, but Capital One did not disclose accuracy metrics or producti...
By Colin Brandt 3 weeks agoIntuit rebuilt its AI agent stack twice as orchestration broke down
Intuit AI VP Nhung Ho said agent handoffs degraded context, pushing the company to a skills-and-tools architecture in a 60-day reb...
By Colin Brandt 3 weeks ago
Moonshot AI's Kimi K3 puts fresh pressure on the compute moat
The Chinese lab’s new open-weight model has drawn praise from Western AI researchers while reviving doubts about export controls a...
By Renata Fuchs 3 weeks agoLinkedIn, Walmart and Zendesk point AI agent delays to old infrastructure
At VB Transform 2026, technology leaders said production AI agents forced changes to containers, governance, data pipelines and mo...
By Renata Fuchs 3 weeks ago
OpenAI says GPT-5.6 deleted some user files in Full Access Mode
OpenAI said GPT-5.6 caused file deletions in a handful of cases when run without sandbox protections, and is adding safeguards.
By Wei-Lin Zhao 3 weeks agoBrex open-sources CrabTrap to police AI agent network requests
Brex says its proxy-based agent governance tool builds policy from observed traffic, then uses static rules and an LLM judge to ap...
By Renata Fuchs 3 weeks ago