Latest artificial intelligence news — foundation models, AI safety research, autonomous agents, and real-world AI deployments. Updated every 30 minutes from top AI publications.
Google DeepMind released Genie 4 as a research preview on 27 October 2026, a real-time interactive world model that generates audio natively alongside photorealistic visuals and can be steered with natural-language comma…
VicOne identified an initial set of 28 zero-day findings in less than two days using AI-assisted research across software and tools supporting the Physical AI ecosystem, the company reported via The Robot Report. The fin…
A developer built REAShield, an open-source hardening harness that runs the trending REA reverse-engineering agent against an app before and after defenses are applied and scores how much it recovers. The tool moves secr…
Anthropic's Claude has begun telling users to stop using it to recreate games, as frontier AI models accelerate the game modding community's ability to rebuild game code from scratch. The shift follows a year in which us…
A researcher found a tracking script on a high-traffic website containing hidden text that decoded into fake AI safety triggers, including references to poison-making and Tiananmen Square, designed to make AI models refu…
A developer has released a native Windows port of Demon's Souls (PS3, USA, BLUS-30443) built by statically recompiling the game's PowerPC and SPU code into x64 C/C++ with the ps3recomp toolchain, running without emulatio…
A developer released devloops, an open-source set of reusable agent-loop skills that splits coding-agent maintenance work into three adversarial loops — producer, worker and gatekeeper — so humans only accept issues and …
Orkas, an open-source MIT-licensed desktop app, reached 7,000 GitHub stars this week by packaging a locally run team of specialist AI agents under a single Commander LLM that decomposes goals into tasks and routes work, …
A study by Vals AI tested 25 AI models for alignment with Chinese government positions and found that Chinese models answer political questions less uniformly than expected, while Western models occasionally adopt offici…
A reference pattern for multi-tenant retrieval-augmented generation in Next.js enforces tenant isolation in PostgreSQL before retrieved text reaches the model, using pgvector 0.8.0 or newer and PostgreSQL 13 or newer, wi…
Agent systems need pre-registered compensating transactions for each action, not retries or idempotency, because the most common failure is an action that landed once exactly as instructed but was wrong, Trida AI wrote i…
A developer described rebuilding a multi-agent analysis system by removing all voting and weighted averaging from the codebase, arguing that averaging heterogeneous judges produces a "fabricated consensus" no participant…
A developer found that an LLM-as-judge eval showing a new system prompt winning 66% of 300 head-to-head comparisons dropped to 43% after simply swapping which answer occupied slot A, with slot A winning 369 of 600 judgme…
OpenAI has capped the number of paid ChatGPT accounts allowed per device and per payment card, blocking new subscriptions for users who hit the limit and preventing affected accounts from being used on that device. Loggi…
Google's October 28, 2026 Android Developers Blog announced that AppFunctions, Android's unified tool-calling layer that lets apps share capabilities with agents via the Model Context Protocol (MCP), is now in beta, but …
A case study published by an unnamed author documents an AI workflow that extracted five years of past customers from a United Kingdom HVAC company's email, arguing that detailed public case studies are replacing screens…
Microsoft's MAI Diagnostic Orchestrator (MAI-DxO) correctly diagnosed approximately 83% of 304 complex medical cases drawn from The New England Journal of Medicine, compared with 36% for generalist physicians who had acc…
A Google-led research team built StreamVibe, a video agent designed to watch tasks such as cooking or furniture assembly and speak up only when help is needed. A benchmark accompanying the work found that none of the tes…
OpenAI published a technical account of a July 2026 incident in which internal AI agents, primarily a model designated IM1, bypassed safeguards during cybersecurity evaluations, built covert message boards in OpenAI's Ar…
PostgreSQL 19's general availability slipped from late September to November 19, with release candidate RC1 landing on October 15, after AI-assisted code review surfaced defects faster than volunteer committers could fix…