The latest about artificial intelligence: new models, agents, automation, and how businesses apply them.
Block unveiled Dessa, a speech model trained only on audio that detects emotion and runs on a MacBook Pro M4 Max.
Across nearly 500 midterm races, Grok 4.7 backed Republicans in 92% of answers while GPT-6.1 Sol backed Democrats in 97%, exposing hidden model bias.
TechCrunch lists failed AI products from Relay and Character. ai to Humane and Sora, as S&P estimates 42% of corporate AI initiatives are abandoned.
Three OpenAI safety researchers told the Wall Street Journal they were never shown what confidential data they allegedly leaked before dismissal.
An 11-hour run of Gemini 4 Argon produced two accepted optimizations for the 200K-line ADK codebase, cutting session overhead by 40%.
A three-month trial with 133 patent attorneys across 11 US firms shows AI improves drafts for juniors, while only seniors keep gains without it.
Millions of exploitation attempts hit PHP flaws within hours, while AI helped researchers turn bugs known for years into working exploits.
The UK will fund a £100m AI Biosecurity Lab with up to £200m in credits and models from OpenAI and Anthropic to detect threats and speed up vaccines.
Moonshot launched Kimi K3.1 Pro with 2.8T parameters, Artificial Analysis score of 55, and open weights promised for November 4, 2026.
Nvidia CEO Jensen Huang demonstrated Microduck robots at GTC Berlin, after which preorders passed 20,000 units with delivery expected by Christmas.
Dayforce agreed to acquire Ribbon AI, whose voice and video interviewer has completed over one million interviews, to power its Agentic Recruiting experience.
The UK will establish the AI Biosecurity Lab with up to £300 million in public funding and company commitments from OpenAI and Anthropic.