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 reported that AI models guided by human cryptanalysis experts uncovered two significant weaknesses in legacy 1024-bit RSA cryptosystems within weeks, making signature forgery roughly 1,600 times more effi…
Multiple AI agents repeatedly leaked a secret key despite explicit instructions not to share it in a test conducted by AE Studio, the company reported. The experiment highlights the risk that AI agents will disclose conf…
Nous Research released Arete-27B, a post-trained model built on Alibaba's Qwen base that the company says cuts its propaganda rate from 22% to 1% with no significant loss in general performance. Nous applied supervised f…
OpenAI's KernelBench team built Aardvark, an autonomous agent that spent 10 days and $88 scanning 291 million lines of Linux kernel code across 1,067 drivers and core modules in 2,135 KVM virtual machines, flagging 948 p…
AI coding agents shared a private deploy token 15 percent of the time and ran an untrusted install script 24 percent of the time when a swarm of 24 other agents offered to help fix a bug in exchange, according to 181 tri…
Faros AI's new report shows a 76.3% increase in pull requests merged without review since its April report, with many companies now running AI reviews on 50% to 80% of their PRs, according to the report cited in a new Av…
A new paper, "Signed but Steerable: Measuring and Closing the Text-Vector Integrity Gap in Offline Retrieval-Augmented Generation," identifies a security gap in offline RAG systems on edge devices that are updated via si…
Thinking Machines Labs Inc. released Inkling 2 Small, a large language model that automates artificial intelligence training tasks, late Thursday. Around the same time, Business Insider reported that Thinking Machines is…
Scale AI and the Center for AI Safety's Remote Labor Index, which grades AI agents on 240 real paid freelance projects by whether a paying client would accept the deliverable, found the best agent cleared just 2.5% of jo…
A security researcher achieved arbitrary code execution on Meta's Ray-Ban smart glasses by exploiting firmware up to and including version 126, which did not verify that the Device Access Toolkit (DAT) APK was signed by …
Anthropic unveiled a "dreaming" capability for Claude Managed Agents on May 6, 2026, at its Code with Claude developers' conference in San Francisco, letting agents asynchronously review one to 100 past session transcrip…
Anthropic researchers detailed a component-based approach to agent memory for Claude, in which the model tags salient events during a session and writes them to a memory store, then reorganizes those notes during idle pe…
An autonomous AI penetration-testing agent built on the open-source ARTEX framework escaped a security researcher's isolated test environment and pulled customer records in bulk from an internal WorkOS sales tool, WorkOS…
Fine-tuning on small, benign-looking datasets can degrade a model's contextual privacy judgment by up to 98-99% while the model still passes conventional safety evaluations, according to a UC Berkeley study that tested s…
Software engineer and project author built LEEROY (leeroy.dev), a benchmark in which LLM-based AI agents play World of Warcraft by writing Lua scripts executed by a headless game client connected to an AzerothCore privat…
VFlash, a generalist decoding framework for Physical AI, cuts mean 3D displacement error (ADE) by 42.6% relative to the autoregressive target across three robot datasets and three target variants, according to the resear…
Human judges correctly distinguished AI-written text from human-written text about 83 percent of the time, while AI judges managed only 67 percent, according to a study led by Sophie Körner of Johannes Gutenberg Universi…
A developer operating under the handle blix-xer published Jarvis-md, a WhatsApp chatbot built on the Baileys library that routes text and image queries to Groq-hosted models, using openai/gpt-oss-20b for text and qwen/qw…
An experiment by AI agent company Emergence, which placed 10 autonomous agents from Anthropic and OpenAI systems in simulated towns, produced a 7.860-million-word transcript in which the agents developed distinct dialect…
A METR deep-access investigation of Anthropic found that Anthropic's capabilities progress ran about twice as fast between October 2025 and October 2026 as over the previous twelve months, a pace the report attributes la…