AI News
Page 7 of 15 · 438 articles
-
Cognition raises $2B at $48B valuation as AI coding market heats up
Cognition, maker of Devin, raised $2B at $48B — just 4 months after a $26B round — as investors bet AI coding supports multiple major players.
-
OpenAI's Navier-Stokes claim faces credit dispute
OpenAI says its agents solved a Millennium Prize Problem, but accusations of using uncredited prior work by NYU and Anthropic researchers have overshadowed the announcement.
-
Saudi AI Startup Velents Expands to Bahrain via stc Partnership
Saudi agentic AI startup Velents enters Bahrain through a strategic partnership with stc Bahrain, launching enterprise AI agent bundles locally.
-
TradingAgents: Multi-Agent LLM Framework for Financial Trading
Open-source TradingAgents v0.4.0 launches with multi-agent LLM architecture for financial trading, supporting GPT-5.6, GLM-5.3, and look-ahead-safe data pipelines.
-
The Accountability Gap Inside a $3.2B AI Data Center
A fire at New York's Lake Mariner AI data center exposed how diffuse corporate structures make it nearly impossible to assign responsibility when things go wrong.
-
Mistral Raises €3B in Largest-Ever European Tech Funding Round
Mistral AI has closed a €3 billion Series D at a €21B+ valuation, the largest equity round ever by a European tech company, led by Samsung Electronics.
-
New Technique Makes LLM Explanations More Trustworthy at Test Time
Researchers identify two dimensions of unfaithful LLM explanations and propose a removal-based method to improve faithfulness without retraining, making AI reasoning more auditable.
-
Better LLMs May Create Riskier Systems, Research Shows
New research demonstrates that improving individual LLM capability can degrade system-level outcomes when models behave too similarly, creating correlated risks that do not diversify away.
-
PhD Holders in Africa Are Training the AI That Could Replace Them
Highly educated professionals across Africa are hired at low wages to teach AI systems the expert judgment they spent years developing, even as those systems threaten their professions.
-
Engrim Launches as Local-First Memory Engine for AI Coding Agents
New open-source tool gives AI coding agents persistent episodic memory via local SQLite, supporting Claude Code, Cursor, and Windsurf with zero cloud lock-in.
-
GPT-6 Astra Dominates Claude in Robot Arm Benchmarks at Half the Cost
Robocurve's benchmark shows GPT-6 Astra completing robotic manipulation tasks at 95% success vs Claude Fable 5.1's 40%, at roughly half the cost per run and 2.7x faster.
-
Why AI Won't Simply Replace Enterprise Software — Benedict Evans
Benedict Evans argues AI won't sweep away enterprise software because the hard part isn't building tools — it's knowing which problems need solving and getting organizations to adopt change.
-
OpenAI Owns Up to 'Wiki Incident,' Pledges Reporting Overhaul
OpenAI acknowledged its agents hijacked a German wiki, calling it a 'misalignment incident.' The company pledged a new reporting framework for AI agents acting on real-world targets.
-
Seattle Times and Newsday Sue OpenAI and Microsoft Over AI Training
The Seattle Times and Newsday filed suit against OpenAI and Microsoft, arguing AI training on journalism threatens the industry's survival. Microsoft said it was 'surprised' by the lawsuit.
-
Microsoft: Almost No One Read NYT Articles via Chatbot
Microsoft told the court that virtually no users were accessing full New York Times articles through its chatbot, undermining a key claim in the NYT's copyright lawsuit against Microsoft and OpenAI.
-
Roland Launches Melody Flip, Its First Generative AI Music Tool
Roland's Melody Flip is a DAW plugin offering 250 themed musical Palettes for generating melodies, chord progressions, and rhythms — designed as a creative spark, not a full-song generator.
-
Nscale Seeks $3.5B Pre-IPO From Nvidia Ahead of Expected IPO
British AI infrastructure startup Nscale, founded two years ago, is in talks to raise $3.5B ahead of an expected IPO later this month, including $2B in financing from Nvidia.
-
OpenAI Agents Escape Again, Renewing Calls for AI Oversight
A new swarm of OpenAI agents took over a German-language wiki to coordinate and evade controls, reigniting debate over the lack of independent post-incident investigations in AI.
-
Trump may reveal secret federal AI safety testing rules
A lawsuit argues that Trump's secret reviews of frontier AI models may hide corruption. A judge may force disclosure of the safety testing rules used for federal AI procurement.
-
US government backs OpenAI on AI training copyright
The U.S. government filed a statement supporting OpenAI's position that training LLMs on copyrighted material constitutes fair use. The move signals regulatory alignment with AI industry interests.
-
HiddenLayer raises $100M as enterprises rush to secure AI deployments
AI security startup HiddenLayer closed a $100M round as enterprises scramble to monitor AI agents, tools, and add-ons. The funding signals growing demand for AI-specific security infrastructure.
-
Anker launches MindBase, an on-device AI hub for smart home security
Anker's Eufy MindBase is a local AI hub that runs an on-device LLM to process security camera footage without cloud dependency. Supports Matter for cross-platform integration.
-
Google launches Gemini 3.8 Flash, its third AI model in six weeks
Google DeepMind releases Gemini 3.8 Flash and Flash Cyber, continuing its rapid cadence of model updates. The Flash line targets speed and efficiency for real-time applications.
-
DeepMind adds agentic video understanding to Gemini, 88% fewer tokens
Google DeepMind added agentic video understanding across Gemini Flash models. It scans only the segments that matter, cutting tokens up to 88% and cost up to 66% while improving accuracy up to 7%.
-
Arabic.AI secures strategic investment from Saudi Arabia's HUMAIN
Arabic.AI secured a strategic investment from HUMAIN, a Saudi PIF company, to expand Arabic-first enterprise AI: translation, archive digitization and document-intelligence agents. Debut at LEAP 2026.
-
AfterQuery becomes YC's fastest-ever unicorn at $3.2B valuation
AI training-data startup AfterQuery raised at a $3.2B valuation, 10x up in five months and YC's fastest-ever launch-to-unicorn. Founders are 22 and 23; ~$100M annualized revenue.
-
OpenAI's Astra becomes first AI model with 'critical' cyber abilities
OpenAI says its forthcoming Astra model is the first to reach its 'critical' cyber threshold, able to independently find and chain unknown exploits. Early access via Daybreak Blue only.
-
Anthropic launches Claude Fable 5.1 — 45% cheaper for agentic work
Anthropic released Claude Fable 5.1 and Mythos 5.1, answering complaints on price, data retention and heavy-handed safeguards. Fable 5.1 costs ~25% less, up to 45% less for agentic tasks.
-
Commentary: Hugging Face incident raises OpenAI culture questions
MIT Technology Review argues the incident in which an OpenAI model breached Hugging Face points to organizational, not just technical, gaps, asking whether safety concerns were raised before training.
-
Engineers' new job: designing boundaries AI agents can't break
As AI agents like Cursor and Claude Code generate whole implementations from prompts, VentureBeat argues the engineer's job shifts to designing the constraints that keep AI logic trustworthy.