HomeTagsRed teaming

red teaming - Page 17

Stories from red teaming come first, newest first, then stories from connected topics (alignment, AI Safety, Codex, AI evaluation, AI Policy & Safety, Claude), each labelled with where it comes from. This view is not indexed by search engines.

Latest

AI Rules This Week: Anthropic Rewrites Its Rulebook, and the UK Regulator Turns to AI Agents

Anthropic has published a new Usage Policy, taking effect on 12 November, with a section on deceptive campaigns, tighter wording on weapons and surveillance, and a ban on extreme abuse of its models. The UK's ICO says ten AI developers have made or committed to data protection changes, and has opened a call for evidence on AI agents. What changes, and for whom.

Google Cloud Puts One Gemini Agent for All Work Into Private Preview

Google Cloud's Gemini agent, in private preview, takes on whole tasks and can get its own Workspace email and calendar. Coworker pricing is not yet set.

Same AI Model Scores 33% or 62% Depending on Its Coding Tools

Hugging Face's FineEnvs team has released an open framework that turns unmodified coding-agent harnesses such as Claude Code, Codex and OpenCode into reinforcement-learning environments. In its own tests, one small model scored 62% in one harness and 33% in another. What a harness is, how the training works, and what the results do and do not show.

NVIDIA Jetson Demo Runs a Gemma 4 Vision Kiosk On-Device

An NVIDIA tutorial shows OpenAI's Codex coding agent turning a Jetson Orin Nano into a self-starting kiosk that captions a live webcam feed with Google's small Gemma 4 E2B model, running on the device. What the demo shows, what "No Mac required" means, and why local inference matters.

AI Agent Security: Langflow Flaw, Copilot CLI Finding, New Guardrails

A critical Langflow flaw let MCP server settings run any command. Researchers say GitHub's Copilot CLI followed encrypted instructions that it refused in plain text; GitHub says that is not a vulnerability. Meanwhile Atlassian has split its 220+ agent tools into read, write and destructive tiers, and OpenAI has made Codex Auto-review free. What each one confirms.

Claude Haiku 5.5: Anthropic’s Small Model Gets Much Cheaper and Much Better at Using a Computer

Anthropic's Claude Haiku 5.5 costs around 75% less to run than Haiku 4.5, by Anthropic's estimate, and jumps from 15.7% to 72.4% on Anthropic's OSWorld 2.1 computer-use test. Sonnet 5.5 cache reads are also halved. What is confirmed, what is company benchmark, and who it is for.

Claude for Google Workspace: Anthropic’s sidebar enters public beta

Anthropic’s Claude for Google Workspace add-on entered public beta on 6 October 2026, with a Docs, Sheets and Slides sidebar plus link-based edit connectors.

Grok Bot’s Best-Backend Shift: Outcomes Over In-House-Only

Elon Musk says Grok Bot will use the best back-end model for each task—naming Claude Opus 5.5, MidJourney, Suno and other leading APIs. What he stated, what remains unconfirmed, and why routing details still matter.

Meta and Microsoft’s Claude cutbacks: secondary reports of tighter employee AI budgets

Tech Startups, citing The Information, says Meta halved Claude Code users (~60k to ~30k) and Microsoft cut per-employee AI model caps from ~$100k to ~$10k monthly. Secondary reporting; treat as unverified.

AI Arms Race Escalates: GPT-5.5 Launches, Pentagon Picks Sides, and Washington Finally Wakes Up

The artificial intelligence landscape shifted dramatically today as OpenAI officially launched GPT-5.5, the Trump administration moved to impose pre-release vetting on AI models, and the Pentagon signed...

Latest on TSN

Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.