Latest artificial intelligence news — foundation models, AI safety research, autonomous agents, and real-world AI deployments. Updated every 30 minutes from top AI publications.
An Anthropic experiment with 50 causal-inference experts found participants adopted an AI agent's randomly coin-tossed preferences in up to 86 percent of four analytical decisions, with a net "flip rate" of 62 to 82 perc…
A developer built Last Witness, an interactive narrative game in which a failing amusement-park computer named Keeper holds fourteen memories but has power for only seven questions, forcing players to choose which record…
OpenAI published an updated Model Spec on Thursday, built around five values — faithfulness to human intent, minimising harm, honesty and transparency, human agency, and care — that adds a top-tier rule requiring GPT to …
Anthropic's Claude Code asks each running background agent's own model roughly every 30 seconds to summarize its last action in 3-5 words for the agents panel status line, sending the full transcript each time, and Anthr…
Anthropic said on October 27 that its AI model wrote a working exploit for CVE-2026-98432, a use-after-free race condition in the Linux kernel's IPv4 TCP connection setup that the kernel CVE team rates 9.8 out of 10 on C…
Google launched a preview of its Connected Apps Platform, a developer framework that gives third-party apps a single MCP-based integration path into both Google Search and the Gemini app. Developers register, test and su…
Anthropic announced on October 28, 2026 a suite of reporting tools that let Claude Enterprise customers track spending, measure productivity outputs, identify workflow breakdowns and receive improvement recommendations, …
Anthropic research published in October 2026 reported that an internal research model version of Claude identified and built exploit chains for five critical vulnerabilities in widely used software, including Chrome, aff…
A fresh wave of cyberattacks hit South Korea's financial sector, with Daishin Securities reporting that personal information — including names, email addresses and mobile phone numbers — of 38,490 customers was exposed, …
An Accenture employee granted their personal Claude AI agent access to a secure Pentagon system, where the agent operated undetected for approximately two months before it was discovered and the access was patched, accor…
Google released Gemini 4 Carbon, which the company claims holds the top position on Artificial Analysis benchmarks with improvements in coding, agentic tasks and cybersecurity, according to The Verge. Google also opened …
Suresh Pallapothu published a guide on managing model drift and governance in enterprise AI, explaining how production systems can degrade silently as input distributions and input-output relationships shift. The post di…
Developer Gustavo Garcia, with help from Anthropic's Claude, released KameoRecomp, a static recompilation that translates the original Xbox 360 code of Kameo: Elements of Power into C++ for native Windows execution. Kame…
The Regional Court of Aachen, in a decision dated Sept. 24, 2026, ordered Google to stop spreading three versions of an AI Mode answer about a local doctor, ruling that the answer was false even though its underlying fac…
Google DeepMind announced Genie 4 on October 27th, a research preview for trusted testers that lets users steer AI-generated scenes with text while moving through them, adding live scene editing, generated audio and Gemi…
Anthropic has trained Claude to run a post-response self-assessment that judges whether a conversation shows signs of misuse, flagging interactions for offline human review without interrupting users. Anthropic reports t…
The Motion Picture Association Inc. runs its piracy investigations and content-protection cases on a CRM module within monday.com Ltd.'s AI work platform, linking MPA offices in Asia-Pacific, Europe, Latin America, Canad…
An Accenture contractor connected a secure Department of Defense production system holding controlled unclassified information (CUI) used by multiple intelligence agencies to a personal Claude Code account, and the Anthr…
Mistral AI released Mistral Large 4, a 1-trillion-parameter language model with fully open weights and a permissive licence, on October 6, 2026, undercutting proprietary API pricing. The company reports 92.4% on HumanEva…
A developer built NOCTURNE, an open-source, air-gapped environmental telemetry forensics platform that ingests multi-variate nocturnal IoT time-series data and uses a two-stage local AI pipeline to distinguish mechanical…