Trust3 AI

This is official account

[Trust3 AI @trust3ai]

@EvanKirstel @TechImpactTV We appreciate the feedback!

[Trust3 AI @trust3ai]

We're all set for #SecureWorld!
Come by booth 460 to grab some awesome swags and talk about #AgentSecurity with us.

#AISecurity

[Trust3 AI @trust3ai]

7am banner assembly at #SecureWorld👷‍♀️
Come find us at booth 460 today and tomorrow!

#AgentSecurity #AIGovernance #AISecurity

[Trust3 AI @trust3ai]

How to Kill and Restore an Agent in Trust3 A https://t.co/rp1cAgrvPG

[Trust3 AI @trust3ai]

CrowdStrike just introduced their AI Kill Switch today. We released the Trust3 AI Kill Switch last week.

Stopping an agent is a different problem from stopping software, because an agent is an identity with standing access to data. You are ultimately revoking reach.

By the time someone wants to stop an agent, it has read from systems and written somewhere.

That makes the stop an access question, and it comes with these requirements.

• Know what you’re reacting to. Every discovered agent carries a Trust Score built from policy compliance, scope adherence, security posture and behavioral baseline. The Trust Score says whether something is wrong, and the activity trace and token spend provides signal of a loop or a hijacked goal.

• Stop the agent without stopping everything else. Agents get provisioned against shared service identities, so revoking the token takes down every other workload using it. That's why we can't simply revoke credentials.

• Carry the record with the action. What called the agent, what it called, what data it reached, who owns it, and the platform state before and after.

• Restore without re-provisioning. If stopping an agent means rebuilding it afterwards, the wait tends to be costly.

We wrote up how the Trust3 AI agent Kill Switch works, check it out here: https://t.co/GLJMlS4rn4

#AgentSecurity #AIGovernance #DataSecurity

[Trust3 AI @trust3ai]

AI agents can access sensitive data through tools like Jira. See how Trust3 AI enforces real-time authorization, blocks unsafe actions, and protects data across agent workflows.

[Trust3 AI @trust3ai]

The next big cybersecurity opportunity is per-action authorization across every data touchpoint—Snowflake, Databricks, SaaS, MCP, and beyond. https://t.co/Sd3uWh4hoI

[Trust3 AI @trust3ai]

We're almost ready #SecureWorld Seattle that's happening this Wednesday and Thursday.

We’ll be talking about agent security - discover, observe, and secure.

Come find us at booth 460 - we'd love to meet fellow agent security enthusiasts!

#AgentSecurity #AISecurity

[Trust3 AI @trust3ai]

We’re at #SecureWorld Seattle this Wednesday and Thursday, booth 460.

We’ll be talking about agent security, from the shadow AI to the data access control for agents.

Come find us at booth 460 - we'd love to meet fellow agent security enthusiasts!

#AgentSecurity #AISecurity

[Trust3 AI @trust3ai]

happy halloweeeennnn @SignalFire

[Trust3 AI @trust3ai]

Happy Halloween!! and Bye Q3 :)

[Trust3 AI @trust3ai]

Happy Halloween @a16z

[Trust3 AI @trust3ai]

Read the full article here: https://t.co/QExFOyJg7E

[Trust3 AI @trust3ai]

New volume of Trust3 AI Labs brought to you by Mark Castillo, our Founding Solutions Architect who's who has worked on hundreds of complex AI security implementations.

This technical deep dive is on agent authorization. How to design a capability token that scopes an agent to one bounded task and keeps the real credentials out of the model’s hands.

It also separates two things that get conflated - short expiry and proof of possession protect a stolen token, while neither one stops the same operation from running twice.

If you are issuing credentials to agents today and you are not sure what happens on a retry after a crash, this one is worth your time.

Full article in the comments.

#AgentSecurity #AgenticAI #AIGovernance

[Trust3 AI @trust3ai]

Our blog goes deeper on what it takes to secure data in the context of agentic AI. Check it out here: https://t.co/2CBI2gh8zO

[Trust3 AI @trust3ai]

Gartner’s new infrastructure cybersecurity on data security report names Trust3 AI among the dedicated agent authorization layers.

The report states that the leading vendors in this market will be securing data for AI agents acting on a person’s behalf.

An agent exists to reach data on a person’s behalf, so agent security that stops at the agent layer is incomplete.

Trust3 AI carries the originating user’s identity and declared purpose through every hop, and enforces one policy at the agent boundary and at the data the agent reaches.

Our blog goes deeper on what it takes to secure data in the context of agentic AI. Check it out in the comments.

#AgentSecurity #DataSecurity #EnterpriseAI

[Trust3 AI @trust3ai]

Here's the full write up: https://t.co/gC1g2TH1H6

[Trust3 AI @trust3ai]

Are you on the #Claude Duet waitlist yet, or already in?

Claude Duet is a proactive agent that runs on its own cloud computer and checks Gmail, Slack, Drive and HubSpot on a schedule, then brings work back to you when it's done.

Very useful but also a little unnerving to us in the security field.

That’s an agent holding an employee’s access and moving data between apps while nobody is in the session. And when a site has no connector, it opens its own browser, which you can’t watch.

We brought Agent DOS to Claude Code this week for the same reason. Both Claude Code and Claude Duet act with a person’s credentials across systems security can’t see into.

We wrote up how to prepare before Duet lands in your org, check it out in the comments.

#AgentSecurity #AIGovernance #EnterpriseAI

[Trust3 AI @trust3ai]

See the full Trust3 AI x Claude Code press releases here: https://t.co/0z9qUol8YE

[Trust3 AI @trust3ai]

Today Trust3 AI is bringing Agent DOS to Claude Code, giving security teams oversight of every Claude Code session and the data/tools/systems it reaches.

“Agent DOS gives security complete oversight, so agent-written code ships under the same scrutiny,” says Balaji Ganesan, Co-founder and CEO of Trust3 AI.

If Claude Code is in your stack, your AI security measures should reach it too.

See the full press release in the comments.

#ClaudeCode #AgentSecurity #AIGovernance