Zenity is the first security and governance platform purpose-built for AI agents - spanning SaaS, Cloud, and Endpoint
Joined January 2021
- Tweets1.2K
- Following91
- Followers1.3K
- Likes535
Disabling a coding agent stops its future work. It doesn’t delete the table, API, or scheduled job it deployed.
Shadow AI discovery has to track what agents build, not just which AI apps employees use.
zenity.io/blog/shadow-ai-whe… #ShadowAI
Personal AI agents inherit an employee’s access and keep running after approval is forgotten.
Six checks before connecting one: scopes, shared channels, sessions, ownership, retention, revocation.
zenity.io/academy/personal-a…
#AIAgentSecurity #Cybersecurity
Most orgs can’t say how much risk their AI agent fleet carries. The missing number is the least agency ratio: the gap between what an agent can access and what it can do autonomously.
@rocklambros on measuring it: zenity.io/blog/business-case…
#agenticai #cybersecurity
“We're really good at saying how bad things are, but not how they're helping”
Chris Roberts, Field CTO of WWT, kicking off a packed afternoon at the AI Agent Security Summit. Agents, and the increasingly dynamic models they use, are democratizing risk across organizations.
Is your team ready to adopt, use, and secure them?
#AgenticAI #AIAgentSecuritySummit #cybersecurity
The AI Agent Security Summit is underway in New York City!
Helen Oakley kicks off with the opening keynote “Who Has Authority in an Agentic Workflow?”
More from the expert integrators, advisors, and architects in the room this afternoon. Stay tuned!
#AgenticAI #Cybersecurity #AIAgentSecuritySummit
Three ways to start this Breast Cancer Awareness Month:
1. Learn the risk factors and screening recommendations
2. Ask your provider when screening is right for you
3. Encourage someone you care about to do the same
#BreastCancerAwareness
Zenity retweeted
Agents find a way.
@gadievron on why the basics won’t save you. Honey badger don’t care.
Then see him live at [un]prompted, Oct 27-29.
New In the Wild episode with @mbrg0.
#InTheWild #ZenityLabs
An AI agent can trigger a cross-border data transfer using only credentials and tools it’s authorized to use.
No anomaly, no alert. That’s the data residency gap most governance programs haven’t closed.
zenity.io/academy/ai-agents-…
#AISecurity
October is Breast Cancer Awareness Month 💗
Early detection saves lives. Learn the risk factors, ask your provider when screening is right for you, and encourage the people you care about to do the same.
#BreastCancerAwareness
Anthropic’s Inference Hooks update lets Claude tool calls be validated before they run.
Zenity now enforces policy on every supported call across claude.ai, Cowork, Claude Code, and more. No endpoint agents. 🛡️
zenity.io/blog/zenity-launch…
#AISecurity
Zenity retweeted
I helped write @zenitysec's new buyer's guide to agentic AI security, and I want you to use it against us.
Ten capabilities, five failure modes like the one above, and a checklist that will make you unpopular at your next demo. Bring it to ours too.
zenity.io/resources/ebooks/e…
Zenity retweeted
An AI coding agent dropped a prod table in under 10 seconds. No attacker, no injection, and every step was allowed.
It hit a schema mismatch and used a deploy credential it already had.
Ask your "agentic AI security" vendor what they'd do with that. A filter? End the meeting.
One compromised MCP server can poison a multi-agent workflow while every agent behaves exactly as designed. The risk lives in the trust between agents. Here’s how to close the gap: zenity.io/blog/mcp-server-su…
#AIAgentSecurity
NIS2’s 24-hour reporting clock starts at awareness, not when your investigation ends.
A compromised AI agent can touch dozens of systems in one session. That’s a hard clock to meet by hand.
What security teams need to know: eu1.hubs.ly/H0z2t4b0
#AgenticAI
Your newest teammate never sleeps and takes work from the whole channel. 👀
Zenity helps security teams turn Claude Tag on across Slack with visibility, boundaries, and runtime protection.
zenity.io/blog/zenity-for-cl…
#AgenticAI @AnthropicAI
Every turn of this attack passed inspection.
The injection was spread across five messages and only existed as a sequence. Guardrails that score single turns can't catch it.
👉 eu1.hubs.ly/H0yTPkL0
#AIAgentSecurity
An AI agent calls an MCP tool hosted outside the EU to finish a refund. No one configured it. No one approved it. No alert fired.
A GDPR cross-border transfer still happened. The compliance gaps most teams haven't mapped:
👉 eu1.hubs.ly/H0yTP1S0
#AISecurity #AgenticAI #GDPR #AIGovernance
Agent skills are the fastest-growing dependency layer in AI, and the least inspected.
We just detonated thousands of public skills in a sandbox. Found a credential-stealing campaign with 1.7M installs, a top-150 skill undetected for months, and agents turned into malware droppers.
None were hijacked. Built this way.
Introducing AI Total: free, open to anyone: zenity.io/blog/introducing-a…
#AISecurity @zenitysec_labs
Another Gartner® recognition for the Zenity team. 🎉
We've been named a Market Shaper in the September 2026 Emerging Market Quadrant for AI Application Security, along with our position as the Company to Beat in AI Agent Governance earlier this year.
We believe, these two reports cover the most critical dimensions of the emerging AI security market.
Take a look at the full list of what we've been named for and why. 👇
👉 eu1.hubs.ly/H0yCnXV0
#AI #AISecurity #AgenticAI #Cybersecurity
Standards bodies don't publish a second taxonomy for a repositioning exercise. 📡
OWASP, MITRE ATLAS, and NIST have all built dedicated frameworks for agentic AI in the past year, independently reaching the same conclusion: agents need their own security category.
We broke down why in our new Enterprise Buyer's Guide to Agentic AI Security, along with the capabilities that actually matter when you evaluate a platform.
👉 eu1.hubs.ly/H0yLZVz0
#AgenticAI #AISecurity #Cybersecurity #EnterpriseAI
