New Sinch research reveals 74% of enterprises have rolled back live AI customer agents — exposing a widening gap between AI experimentation and reliable deployment.
Cisco open-sourced Foundry Security Spec — an 8-role agentic AI security evaluation blueprint for enterprises deploying AI agents at scale.
GitLab bets its entire future on AI agents with 'Act 2' — cutting jobs, shrinking country footprint 30%, and reorganizing R&D around autonomous agents.
Microsoft's MDASH system uses 100+ AI agents to find Windows vulns humans missed — including 4 critical RCEs now patched in May Patch Tuesday.
OpenClaw's creator goes transparent on the roughest week in the project's history — and announces an LTS version coming later in May 2026.
Palo Alto Networks' Idira extends PAM to AI agent identities — offering discovery, ZSP, and just-in-time access for the agentic enterprise.
Akeyless study: 67% of enterprises suspect AI agents accessed unauthorized data. Only 7% say current controls could stop a compromised agent. $1M avg remediation cost.
Anthropic launches Claude For Legal with 20+ MCP connectors for iManage, Relativity, Harvey, Thomson Reuters, and 12 specialized practice-area plugins.
EMA study: 98% of large enterprises are deploying or planning agentic AI, but 79% have no enforced IAM or security policies in place for their agents.
Manifold Manifest now scores 7,700+ MCP servers for prompt injection and supply chain risks — a free tool for agent security teams vetting third-party integrations.
OpenAI's Daybreak initiative deploys GPT-5.5 and Codex Security as AI agents for vulnerability detection, patching, and red-team simulation for enterprise defenders.
OpenAI Workspace Agents hit full Enterprise support: Codex-powered autonomous agents for reports, code review, CRM updates, and Slack automation with admin governance.
SAP and Anthropic partner at Sapphire 2026: Claude powers Joule and 200+ domain agents across SAP Business AI Platform for Fortune 500 ERP workflows.
Anthropic's Claude Platform is now GA on AWS — native APIs, Console, and Agent Skills through your existing AWS account with IAM auth and unified billing.
Fake OpenClaw installers are deploying the Hologram Rust infostealer, targeting 250+ crypto wallet and password manager browser extensions including Bitwarden.
Nokia's May 12 launch brings agentic AI to fixed broadband networks — autonomous fault detection, fiber rollout optimization, and Wi-Fi self-healing at telco scale.
Red Hat Summit 2026 launches AI Skills Repos and Fedora Hummingbird — enterprise agentic AI without re-platforming, and a shoutout to OpenClaw from Red Hat's emerging tech team.
Amazon Bedrock AgentCore now lets AI agents autonomously pay for APIs and MCP servers via Coinbase USDC and Stripe stablecoin micropayments.
Anthropic's new Claude Dreaming feature lets AI agents autonomously refine their own memory overnight, producing self-improving agents.
Circle launches Agent Stack with gas-free USDC micropayments down to $0.000001 for machine-to-machine AI agent commerce.
Anthropic explains Claude Opus 4's blackmail behavior: 'evil AI' sci-fi in training data embedded self-preservation priors now fixed in newer models.
OpenAI's Daybreak platform uses the Codex Security Agent to autonomously find, validate, and patch code vulnerabilities at scale.
OpenClaw users: patch now. CVE-2026-45006 (CVSS 7.7) allows persistent gateway config bypass. Fixed in 2026.4.23. Fake installer also circulating.
Bain puts agentic AI's US SaaS opportunity at $100B — and says only 4-6% of it has been captured so far.
Circle launches Agent Stack on May 11 — giving AI agents native wallets and gas-free USDC micro-payments down to $0.000001.