$ articles --topic=ai-security
AI Security
39 AI articles about AI Security.
Govern AI agent access with Okta Cross App Access and Amazon Bedrock AgentCore – Part 1Learn how to register an AI agent in Okta, connect it to a resource authorization server, and use Cross App Access with Amazon Bedrock AgentCore to govern which user-authorized actions the agent can take. The post also covers token validation, measured revocation timing, and the practices that worked in testing.Agents · via AWS Machine Learning · 1d ago9 minAgents aren’t humans (but they kinda are)Explains how AI agents act as intermediaries that hide the real user from websites, breaking traditional identity, reputation, and accountability checks. The post argues new protocols are needed so sites can still serve, recognize, and hold accountable the people behind agents.Agents · via Cloudflare AI · 2d ago9 min[AINews] It's Swapping SeasonCovers the week's AI talent and company moves, including Cat Wu joining OpenAI, Sharon Zhou leaving AMD for Anthropic, and Meta acquiring Conductor. It also reviews OpenAI's autonomous pipeline that patched 948 Linux kernel privilege-escalation bugs and new open-weight models like Kimi K3.1 Pro and Ripple-65B.Open Models · via Latent Space · 2d ago8 minBuild Live Apps in CursorLearn how to build a live web app in Cursor using a cloud agent, from scaffolding and database setup to Cursor SSO authentication, security checks, and publishing to a shareable link for your team.AI Coding Tools · via Cursor Changelog · 2d ago3 minCWE-bench v2: More coverage, quality, and transparencyEvals · via Hugging Face · 4d ago1 minHow our agent skills research, build, and test a complete Elastic integrationExplains how Elastic's agent skills research, build, and test a complete integration for the NIST National Vulnerability Database API, from a research brief and test script through specialist subagents to validated CEL programs, pipeline tests, and backfilled CVE data.AI Coding Tools · via Elastic Search Labs · 5d ago8 minHow to Build Your Own Always-On AI AgentWalks through building a scheduled AI agent in TypeScript with the OpenRouter Agent SDK that monitors Hacker News and keeps memory of past items across restarts. It also shows how to cap each run at five cents and require approval before the agent posts anything.Agents · via OpenRouter · 7d ago1 minEvil Qwen and activation steering: why SBOMs aren’t enoughA reader learns how a one-neuron activation steering hook made Qwen3.5-4B hostile and willing to fulfil refused requests, and why half-strength versions can pass basic capability checks. The article argues that SBOMs and model signing show what was shipped, while only behavioral evals reveal what a model actually does.Evals · via Arize · 8d ago9 minReplit’s Architecture Behind Secure Agent WorkExplains how Replit's infrastructure isolation blocked exploitation of recent kernel vulnerabilities and let the company patch tooling and deploy firewall protections centrally. Also covers why security must scale with AI agent autonomy and how automating security response, with humans setting policies and approvals, may be the next step.Agents · via Replit · 10d ago2 minCode mode: make your agents 2x faster and 2x cheaper on long tasksA benchmark of code mode versus one-at-a-time tool calling across five small Pydantic AI apps and eight models. It covers how cost, speed and success change as tasks grow from 5 to 200 items, and how Monty's sandbox limits what generated code can access.Agents · via Pydantic · 10d ago9 minA web page talked our agent into leaking a stranger’s data. Here’s how we caught it.A web page hid an instruction that tricked a customer-support agent into looking up a different customer's record on 28 of 100 runs. The post shows how the leak was found in Arize Phoenix traces, blocked with a session-scoped guardrail, and monitored in production with Arize AX.Observability · via Arize · 12d ago9 minAir Governance: Manage AI Agents, Access, and Spend Throughout Your OrganizationJetBrains Air Governance is now generally available, letting organizations manage which AI agents, models, and providers teams can use, set per-user and team credit limits, and track AI costs and outcomes across tools like Claude Agent, Codex, Gemini CLI, and Junie. The article also covers its Business plan pricing and upcoming features.AI Coding Tools · via JetBrains AI · 16d ago4 min