Zenity

October is Breast Cancer Awareness Month 💗 In tech, it’s easy to stay focused on what’s next. The pace is fast, and our own health can slip down the list. This month is a good reminder to make time for it. Early detection can save lives, and staying informed is an important first step. Three ways to start: → Learn about risk factors and screening recommendations → Ask your provider when screening is right for you → Encourage the people you care about to do the same Take a moment this month for yourself and for someone you care about. #BreastCancerAwareness

Europe has very few homegrown general models. So what does that mean for our ability to control AI? Keren Katz and John Sotiropoulos sat down at the AI Agent Security Summit in London last week. Their take: sovereignty, resilience and long-term security have to be tackled together. Great regulation alone won’t get you there. We can’t control what politicians decide. We can control our security baseline. Focus on the fundamentals. Want more conversations like this? Join us at the AI Agent Security Summit in NYC on October 21. https://lnkd.in/emcRQ_Ku #AIAgentSecuritySummit #AgenticAI #AISecurity

Today, Zenity is launching Universal Runtime Control for Claude Enterprise, so policy applies to each action across: 🔎 claude.ai, Cowork, and Claude Code ⚙️ Claude in Chrome and Claude for Microsoft 365 ⛔ Claude Tag in Slack Security teams shouldn’t have to choose between approving Claude and controlling it. 🛡️ Anthropic’s Inference Hooks update now holds Claude’s proposed tool calls for an allow-or-deny verdict before they run. One configuration. Nothing to install on endpoints. That means teams can approve write-capable connectors, MCP servers, and wider Claude Code rollouts, then block the unauthorized delete, the unverified server, or the action driven by prompt injection before it executes. Intent is not control. 👉 https://lnkd.in/efjfXite #AISecurity #AgenticAI #Cybersecurity #AIAgentSecurity Snir Hassidim Anthropic

You're invited to KaiZen on November 10! 🎌 At Zenity's inaugural event in Tokyo, join security pioneers, AI strategists, and our executive leadership for a day breaking down the state of AI security in local enterprises: and how to adapt when agent rollout starts accelerating. See the agenda and register now! 👇 https://lnkd.in/evRKGPP4 #AISecurity #AgenticAI #Cybersecurity Dai Fujimoto

Finance is building apps. Procurement is building apps. And security is still expected to police all of it. Keren Katz and Gary Hayslip sat down at the AI Agent Security Summit in London last week. Their take: people outside traditional IT are now building with code, and the pressure to save money only speeds it up. It’s moving fast, and it’s fascinating to watch. Watch the full interview: https://lnkd.in/ecin2e85 Want more conversations like this? Join us at the AI Agent Security Summit in NYC on October 21. Sign up here: https://lnkd.in/emcRQ_Ku #AIAgentSecuritySummit #AgenticAI #cybersecurity

No agent in the workflow misbehaved. The backdoor shipped anyway. 🔓 Picture three AI agents: code review, deployment, and monitoring. All coordinating through one shared MCP server. An attacker poisons a dependency inside that server. The deployment agent ships a backdoor and reports a clean release. The monitoring agent trusts the report. Every agent did exactly what it was designed to do. The compromise lives in the relationships between them. 🔗 That’s why watching agents one by one isn’t enough. MCP servers are supply chain dependencies, and every agent you add creates another edge in the trust graph. In our latest blog, we break down how to inventory the servers your agents connect to, verify their integrity, and detect anomalous instruction flows between agents: 👉 https://lnkd.in/eK4tVxFT #AISecurity #AgenticAI #AIAgentSecurity Rock Lambros

NIS2 gives you 24 hours to notify your CSIRT of a significant incident. The clock starts when you become aware, not when the investigation ends. ⏱️ Now apply that to AI agents. A single compromised agent can traverse multiple systems, invoke dozens of tool calls, and touch data across the enterprise in one session. Reconstructing that by hand takes far longer than a day. “Awareness” is also murky for agentic incidents. Regulators haven’t said whether it starts at the first anomalous session or when the full scope is confirmed. The safer reading: the moment your detection surfaces a credible signal. 🔎 That puts real pressure on three things: → Real-time detection (Article 23 reporting timelines) → MCP server and supply chain visibility (Article 21) → Board-level reporting (Article 20) Detection after exfiltration is not security. 🛡️ Our latest blog breaks down what NIS2 means for agentic AI security, with an evaluation checklist you can use today: 👉 https://lnkd.in/dtK9xYue #AISecurity #AgenticAI #Cybersecurity #NIS2

No region has moved faster on binding AI regulation. 📘 EU AI Act. GDPR. NIS2. DORA. Together, they create compliance obligations for agentic AI that most other markets simply don't have yet, and an agent that can be manipulated via prompt injection or that operates without audit-ready logging isn't just insecure. It's potentially non-compliant. Our new EMEA Market Addendum maps each regulation to the specific security capability it requires, from conformity documentation to cross-border data flow monitoring. Download it and you get the full Enterprise Buyer's Guide to Agentic AI Security too. 👉 https://lnkd.in/egkucbT3 #AISecurity #AIGovernance #Cybersecurity #AgenticAI

AIエージェントは、実際に何にアクセスできるのでしょうか？ 7月に東京で開催された「CTC DISCOVER SECURITY」では、ZenityのJustin VitaとChen Yangが登壇し、ライブデモを通じてこの問いにお答えしました。 デモでは、エージェントが利用するツール、データ、接続先を可視化し、セキュリティリスクを明らかにしました。さらに、その結果をもとに対策の優先順位を決め、エージェントの行動を企業のポリシーに沿って制御するルールを適用する方法をご紹介しました。 イベントでの議論をブログシリーズで広く紹介してくださったCTCの皆さまに、感謝申し上げます。 → CTCのAIセキュリティ戦略とロードマップ： https://lnkd.in/eSY94XpX → Zenityの講演を紹介したレポート： https://lnkd.in/enmh58rW 10月16日（金）に東京で開催される「CTC DISCOVER 2026」では、Microsoft Copilot、OpenAI Codex、Anthropic Claude CodeなどのAIエージェントを、Zenityがどのように保護するのかをご紹介します。 AIエージェントの活用が試験的な段階から本格導入へと広がる中、アクセス先と行動を把握し、企業のポリシーに沿って制御する方法を、ぜひ会場でご覧ください。 #AIセキュリティ #AIエージェント ----- What can your AI agents actually access? In July, Zenity’s Justin VITA and Chen Yang answered that question through a live demo at CTC DISCOVER SECURITY in Tokyo. Our demo showed how mapping an agent’s tools, data, and connections reveals security exposure. We also demonstrated how teams can use those findings to prioritize remediation and enforce boundaries that keep agent behavior aligned with business policy. Thank you to CTC for sharing the event’s discussions with a wider audience through their blog series: → CTC’s AI security strategy and roadmap: https://lnkd.in/eSY94XpX → Report featuring Zenity’s presentation: https://lnkd.in/enmh58rW Join us on Friday, October 16, at CTC DISCOVER 2026 in Tokyo to learn how Zenity helps secure AI agents such as Microsoft Copilot, OpenAI Codex, and Anthropic Claude Code. As adoption expands from experimentation to broader deployment, see how to understand what these agents can access, monitor their behavior, and enforce controls aligned with your business policies. #AIsecurity #AIagents Dai Fujimoto イベントの詳細はこちら / Event details: https://lnkd.in/eQ43QBby

62.1% of people in Japan use AI individually. Only 13.1% of businesses have integrated it into real business processes. 📊 That gap is about to close. And once AI agents access real data and take action, the risk changes with it. So Zenity is making its commitment to Japan local. Dai Fujimoto joins as Japan Country Manager, bringing 20+ years at NTT and leadership roles at FireEye and SecurityScorecard. 🤝 His take: now is the time to build governance in from the start, not bolt it on afterward. Meet the team in Tokyo: 📍 CTC Discover 2026, October 16, Booth 3 📍 KaiZen, November 10, ANA InterContinental Tokyo 📍 CODE BLUE 2026, November 17 to 18 👉 https://lnkd.in/gfeM44nq #AISecurity #AgenticAI #Cybersecurity #AIGovernance