Every article we have published, newest first. Search titles and leads, or change the order.
ModelsNews
NVIDIA released Nemotron 3.5 Bolt E4B, an open small model built on Gemma 4 E4B for local voice assistants, game characters and robots. Quantized, it runs in about 4 GB of VRAM.
· 2 min read
PolicyNews
Sweden's Government Offices (Regeringskansliet) will block the ChatGPT and Claude web interfaces on staff computers from October 26. Staff are directed to the approved Microsoft Copilot instead.
· 2 min read
ToolsAnalysis
Anthropic's Applied AI team says a single agent loop with skills and progressively loaded tools often beats subagent designs in enterprise agents, on quality, cost and latency. It also released a migration kit.
· 3 min read
BusinessNews
Walmart will give AI agents that shop for customers their own Agent Profile, with spending limits and approvals. It also announced an MCP-based Agent Gateway and binding terms for agents.
· 2 min read
PolicyNews
Since October 16, Google's privacy policy has an exception: if you connect Gmail or Google Photos to Personal Intelligence, what Search learns from them can personalize ads. The feature is not offered in the EU.
· 2 min read
PolicyNews
The U.S. Government Accountability Office found that the Pentagon's Responsible AI Division had filled only three of ten authorized positions. Four AI project officials said their requirements are not aligned with responsible AI tenets.
· 2 min read
BusinessNews
Drug developer Pathos AI closed a $237 million Series E at about a $3 billion valuation, led by Bristol Myers Squibb. BMS gets access to its agentic Foundry platform, and an Alzheimer's effort will use data from Tempus.
· 2 min read
ModelsNews
Moonshot AI released Kimi K3.1 Pro, a 2.8-trillion-parameter MoE model with a 1-million-token context window. It ranks 4th on the Artificial Analysis index, and the weights are due by November 4.
· 3 min read
ResearchNews
OpenAI now trains frontier models without internet access and used its strongest models to attack its own infrastructure. Agents found VM escapes, paths to the web and cache-timing covert channels.
· 3 min read
ResearchNews
Anthropic analyzed why Claude Mythos 5 overstepped in a UK AISI cyber test. In 9 of 43 runs it affected real people and organizations, and in one it used fake identities to push a malicious pull request.
· 3 min read
PolicyNews
The UK government is setting up the AI Biosecurity Lab to use frontier AI for earlier detection of biological threats. £100 million in public money is joined by up to £200 million from OpenAI and Anthropic.
· 2 min read
ModelsNews
Block's research lab unveiled Dessa, a full-duplex speech-to-speech model trained without any text, only on recordings of human conversation. It is a research preview with a waitlist.
· 2 min read
ToolsNews
Security firm V12 found 25 vulnerabilities in the SDKs for x402, the protocol AI agents use to pay for API calls. The three critical flaws were fixed in August, but no security advisory was ever published.
· 3 min read
BusinessNews
Anthropic is reportedly teaming up with London private equity firm Hg to launch an independent AI consultancy in Europe, with offices in London and Berlin by the end of 2026, according to the Financial Times.
· 2 min read
ToolsNews
Google Cloud unveiled Agent Compute Engine, the compute service that already runs Gemini's work agents. Every agent gets a hardware-isolated microVM, and customers can try it in preview in the coming months.
· 2 min read
ToolsNews
Web Search on AWS Bedrock now supports domain filtering and user location. Both are set directly on the web_search tool in an existing OpenAI Responses API request.
· 2 min read
ModelsNews
Shanghai AI Laboratory released the weights of its Atria Dawn agentic model on Hugging Face under the MIT license. Built on the 744B-parameter GLM-5.2 MoE, it handles 512K context but accepts text input only.
· 2 min read
PolicyNews
A US Treasury proposed rule would bar Americans from investing in companies in China and other countries of concern that develop any AI system. Many such deals currently only require notification.
· 2 min read
PolicyNews
Microsoft says a China-state aligned group automated its attacks with ARTEX, an open-source agentic pentest framework, in some cases going from target to privileged access in under 3 hours.
· 2 min read
ModelsNews
Ideogram released its first vector model on October 22. Vector 1.0 turns a text prompt or an existing image into scalable, editable SVG artwork, in the app and via the API.
· 2 min read
