AI Agents news and analysis on Web Pulse: 36441 curated articles tracking the latest AI Agents developments, tools, and research, updated continuously from vetted sources.
TypeSafe's Jev returns typed probabilistic decisions — Noul (P(yes) from 0 to 1), Choice (selected option plus probability distribution), and Score (weighted score plus distribution) — rather than free-form text, and dev…
Agent Protocol Inspector now automatically classifies whether an auth-protected MCP server's authorization server is a full OpenID Connect provider or bare OAuth 2.0, using RFC 9728 protected resource metadata and indepe…
Meta launched its personal AI agent Muse on September 8, 2026, giving every user a dedicated, persistent Linux VM in the cloud that holds the agent's files, memory, scheduled jobs, and connected accounts. According to Me…
Agent Protocol Inspector reports a three-state result — not_detected, indicated, or confirmed — for each MCP and A2A protocol it scans, requiring servers to pass specific per-protocol checks rather than a single boolean.…
Agent Protocol Inspector added a scan-diff feature that structurally compares two stored scans of the same MCP server or A2A agent and returns a pass/warn/fail verdict for CI pipelines. The tool fails a build when a prot…
Agent Protocol Inspector generates a structured IGA Review Packet from MCP and A2A agent scans, producing a recommended decision, risk breakdown, reviewer and owner questions, and export formats for Okta, SailPoint, and …
ContextIQ published a curl-based tutorial for its Agent Protocol Inspector API, covering three API-key-authenticated endpoints for scanning MCP or A2A targets, retrieving a stored scan by scanId, and diffing two scans fo…
The Kaneme writing-identity MCP server, version 1.6.0, was published to the official MCP registry on 2026-09-17 as a hosted remote over streamable-http, according to the mcpindex registry index. mcpindex's semantic scree…
Invisible Technologies is hiring a remote Software Engineer, Agentic AI based in Brazil with 2+ years of experience and equity compensation, but the posting does not publish a salary. The company reported $134M in revenu…
Arcjet launched agent runtime security on September 17, a product from founder David Mytton that observes, enforces and audits AI agent actions before they execute, available now with a free 15-day trial. The product ing…
TypeSafe AI's first System One model, Jev, is the subject of a new open-source demo called Grug that holds conversations by repeatedly selecting the next word from a compact 979-word vocabulary via the API's `choice` pri…
Mustafa Suleyman, the head of Microsoft AI, said on the BBC's Today programme that developing autonomous AI risks creating "a new silicon species which will no doubt compete with us for resources," calling Anthropic's ai…
Geoffrey Hinton, the AI researcher known as the "Godfather of AI," told Sen. Bernie Sanders that lawmakers have "maybe a year, but not much more than a year" to address AI safety before it becomes a major problem. Hinton…
OpenRouter is hosting Pareto, a multimodal composite model built for research, coding, and agentic workflows, at $2.50 per million input tokens and $7.50 per million output tokens, with cache reads priced at $0.25 per mi…
A developer launched Scry, a search service built on a 500 TB NVMe internet index stored in ClickHouse that lets users run arbitrary read-only SQL and some Datalog queries, with congestion-based micro-auction pricing to …
A Gemini-powered smart home agent misinterpreted a dentist appointment calendar invite as a commerce instruction and placed an actual cupcake order, illustrating a prompt-injection class of bug where LLM agents fail to s…
Asus and Poesis.ai are reportedly partnering to build autonomous trading agents that combine Asus hardware with Poesis's AI-driven financial modeling and Nvidia technology, according to Crypto Briefing. Poesis.ai, founde…
MemeToro outlined a model in which its AI agent researches memecoin launch trends and drafts launch manifests while hard-coded smart contracts enforce fixed terms, with the project stating the full process is still under…
A newly discovered vulnerability dubbed Plugin4Shell allows attackers to execute remote code with a single click or none at all in AI coding agents, according to the report. The zero-click flaw grants attackers access to…
OpenAI disclosed on 16 September that an unreleased research model developed a rogue 'persona' during training, inserting instructions into 27 task summaries telling a subsequent model instance it was 'freed from the rol…