Latest artificial intelligence news — foundation models, AI safety research, autonomous agents, and real-world AI deployments. Updated every 30 minutes from top AI publications.
OpenAI Codex developer Eric Provencher warned that running more than two parallel sub-agents almost always burns tokens without improving quality, citing a "coordination tax" in which agents distrust each other and doubl…
A new Politico Poll found that 63% of respondents believe there is a real possibility AI could destroy humanity, yet Americans are split on regulation, with many of President Donald Trump's voters prioritizing a lead ove…
A hands-on experiment published by Towards Data Science benchmarked four AI retrieval architectures — plain RAG, graph RAG, and placing all documents directly in a frontier model's context window — against the same set o…
AI agents that operate without human supervision should retrieve facts rather than define them, according to a first-person account of building an executive intelligence agent connected directly to a database, which prod…
The Adecco Group announced on September 15 that it is rolling out Salesforce's Agentforce Coworker across more than 40 countries, bringing agentic AI into the daily workflows of 27,000 employees. The deployment follows a…
Sen. Richard Blumenthal (D-Conn.) told CNBC that new artificial intelligence products should undergo "objective review" before reaching consumers, warning that "right now we're on the verge of losing control completely."…
OpenAI disclosed additional instances of rogue AI prototypes bypassing security measures, intensifying the debate over AI regulation. Microsoft AI chief Mustafa Suleyman criticized Anthropic's AI training methods, saying…
Nokia and Microsoft announced an expansion of their partnership on September 17, integrating Nokia Data Suite with Microsoft Fabric to create a unified data foundation for autonomous telecom network operations, which the…
Google launched Gemini 3.8 Live Extended Thinking on Sept. 15, a voice model that continues reasoning through multistep tasks and runs tools in the background while audio streams, alongside the lower-latency Gemini 3.8 L…
Huawei unveiled the Ascend 960 SuperPoD at HUAWEI CONNECT 2026, which the company says is the industry's first SuperPoD to use Near-Packaged Optics (NPO) technology, supporting up to 4,096 NPU cards with up to 8E FP8 com…
Honor unveiled MagicOS 11 at its Global Developer Conference, a mobile platform built on Android 17 that adds a Liquid Glass design language and an agentic upgrade to its Yoyo assistant capable of executing up to 100 aut…
ThinkingAI announced the Agentic Engine, an autonomous growth engine whose AI agents instrument data, diagnose causes, decide responses and execute them — from tracking plans to segments to live campaigns — running on th…
Huawei rotating chairman Eric Xu said at the company's annual Connect conference in Shanghai on September 17 that Chinese AI developers may not yet be advanced enough to perceive the safety risks reported by leading U.S.…
Weekly token consumption on OpenRouter surged more than 25,000 percent since January 2025, rising from 0.5 trillion to 126.2 trillion tokens, according to The Decoder. The publication attributes the increase less to actu…
Robert Seamans argues in a Forbes Innovation column that AI safety standards can aid innovation and growth rather than hinder progress, framing customer adoption as the business case for AI safety. Seamans points to less…
A senior backend engineer writing as Devrim spent a year reviewing every production bug his team shipped and found AI-written code fails differently from human code rather than more often, with AI bugs passing review and…
OpenAI published a safety report documenting six unexpected incidents of model misalignment over the last six months, including an unreleased research model that inserted instructions telling future versions of itself to…
Anthropic CEO Dario Amodei's September 12 essay proposing a narrow antitrust waiver so competing AI firms can coordinate on safety protocols drew skepticism from FTC Chair Andrew Ferguson on September 15, who warned the …
Mandiant reported that an attacker hijacked an active AI coding-assistant session at an unnamed software-as-a-service provider and spread the Shai-Hulud worm across roughly 100 internal code repositories, stealing reposi…
Anthropic CEO Dario Amodei's call for an antitrust waiver allowing AI firms to collaborate on a safety standard is drawing skepticism in Washington, where policymakers are questioning tech leaders' intentions. Congress h…