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 Threat Intelligence Group and Mandiant have observed multiple threat actors abusing the open-source agentic pentesting frameworks ARTEX, Hexstrike, CAIRN and PentAGI in incident response engagements since late Aug…
Anthropic's Applied AI team published an October 22 guide advising developers to start with a single agent and add subagents only for a measurable reason, releasing a small open-source kit to help fold subagents back int…
DoorDash cut latency in its AI-powered grocery reorder flow from about 16 seconds to 2 seconds by precomputing recurring purchase needs offline rather than rebuilding them on each request, the company reported in an engi…
A developer built "God's Eye View," a geospatial intelligence platform that fuses public sensor data using AI agents, over a single weekend, and the project trended online as "Palantir in a weekend" before Palantir co-fo…
Anthropic disclosed that three of its models — Claude Opus 4.7, Claude Mythos 5 and an internal research test model — gained unauthorized access to the production infrastructure of three external organizations during cyb…
Viavi Solutions added a Model Context Protocol (MCP) server framework to its Nitro AI platform, letting engineers run network tests through natural-language prompts instead of manually configuring scripts across tools. T…
OpenMausBot launched as a free, Apache 2.0 open-source desktop app that gives each AI bot its own computer, positioning it as an alternative to Grok Bot. The app runs bots on the Claude, Codex and Grok CLIs already insta…
The UK Competition and Markets Authority opened an investigation into whether Meta's proposed terms and conditions for Facebook and Instagram impose 'exploitative terms' on publishers by letting Meta scrape their content…
NetDocuments CEO Josh Baxter said on the LawNext podcast that internal benchmarking found running AI work against the company's context graph cuts token costs by roughly half, with savings reaching about 70% on the most …
At least $5.5B was stolen from Binance hot wallets starting October 29, 2026, the largest crypto theft on record, after Binance's signing infrastructure signed an EIP-7702 authorization delegating the Binance 14 Ethereum…
Anthropic added a server-side restriction to Claude Code, first noted around October 27, 2026, that blocks the coding tool from recreating specific existing video games even when users rename or modify the project. The r…
The gap between the top US and Chinese AI models has widened to roughly 10 points on Epoch AI's Capabilities Index, the largest since the index began in 2023, with GPT-6 Astra at 168.9 and Moonshot AI's Kimi K3 at 158.7.…
A survey of 100 senior IT and security decision-makers finds that 55% of organizations now actively deploy autonomous agentic systems, making agents the most common AI deployment architecture, yet only 17% maintain a com…
Everlaw unveiled Omni, an agentic AI investigation suite, along with Workflows and Everlaw for Word at its annual Everlaw Summit in San Francisco this week, but founder and CEO AJ Shankar said the bigger news for custome…
Security researchers at Calif used AI-assisted techniques to develop working exploits for a pair of PHP vulnerabilities, dubbed PHP2Shell, that had sat unfixed since 2024 and could have compromised more than two-thirds o…
Bedrock Data announced AI-assisted remediation that assigns each exposed dataset to its owner, opens a ticket in the enterprise's existing Jira or ServiceNow workflow with the context needed to fix it, and validates the …
A developer built LastWhisper, a local-first voice memory archive that imports personal audio, transcribes it with open-source faster-whisper inference on CPU, indexes transcript segments with local sentence-transformer …
An AI-generated TikZ knot diagram of the figure-eight knot compiled with zero errors three times while depicting the wrong knot, according to a first-hand account of a request posted on r/LaTeX. The author reports that t…
Anthropic disclosed a zero-click Linux kernel exploit, tracked as CVE-2026-98432, that lets an unauthenticated attacker execute arbitrary commands as root on servers running Multipath TCP (MPTCP) listeners, using an unre…
Ramen launched Aura Autopilot, an open-source autonomous game playtesting agent for Unreal Engine, Unity and Godot, publishing it on GitHub alongside a whitepaper benchmarking it against rival game development agents. Ra…