DecisionGuard

DecisionGuard gates agent & pipeline changes before execution. Deterministic verdicts, audit-grade evidence, no hallucinations:

[DecisionGuard @DecisionGuard]

Most incidents don’t start as incidents.
They start as unchecked decisions.

We just launched a new DecisionGuard landing page + a 30-second trial.

Govern decisions before systems act.

https://t.co/TvRYVeUMlj

#DecisionGovernance #AgenticAI #SecurityByDesign

[DecisionGuard @DecisionGuard]

Switch AI models. Keep your governance.
Claude, GPT or in-house, calling a SaaS API, a server or your own tools: same rules for every action.
DecisionGuard governs the action, not the model. https://t.co/TvRYVeVkaR #AIAgents

[DecisionGuard @DecisionGuard]

@tiboadriana https://t.co/QJpIZjQWa8

[DecisionGuard @DecisionGuard]

Agreed, Keith. The interesting part is what happens between those control points.

Different platforms bring their own context and signals, but you still need a consistent way to evaluate actions, apply policy and capture evidence.

That's what we're building with DecisionGuard. Keep the decision authority independent, while allowing the underlying technology to evolve.

[DecisionGuard @DecisionGuard]

One authority. Many control points.

DG Edge · Browser Authority + Credential Vault · ToolHub · Workplace (upcoming) · MCP & API directories, SDKs.

Switch models or AI vendors without rebuilding your governance.
We sign actions, not sessions.
https://t.co/TvRYVeVkaR

[DecisionGuard @DecisionGuard]

Compliance evidence shouldn't wait for audit season.

Every governed agent review → verdict → GOV codes, in real time → mapped to ISO 27001, NIS2, DORA, EU AI Act, SOC 2, NIST CSF, NIST AI RMF, HIPAA and PCI DSS.

Control health live. Evidence exportable.
https://t.co/tRrzwMCiDu

[DecisionGuard @DecisionGuard]

Cloud agents can't reach your firewalls, switches or OT.
DecisionGuard's on-prem hub relay connects over an encrypted tunnel, with nothing internal exposed, and every action is still checked before it runs.
We sign actions, not sessions. https://t.co/tRrzwMCiDu #AIAgents #OTSecurity

[DecisionGuard @DecisionGuard]

Think VDI, but for AI agents.

The vault fills the password. The AI never sees it.
Every sign-in step logged.
A human takes over? The agent pauses.

VDI governs who gets a desktop. DecisionGuard governs every action inside it.
Coming soon. https://t.co/TvRYVeUMlj

[DecisionGuard @DecisionGuard]

@MartinGTobias ZeroTrust Execution

[DecisionGuard @DecisionGuard]

Your AI agent has a valid session. But what data is inside its next action?
DecisionGuard inspects every action for PII and sensitive data at execution time, and decides before it runs: allow, block or require approval.
We sign actions, not sessions. https://t.co/tRrzwMCiDu #AIAgents #AIGovernance

[DecisionGuard @DecisionGuard]

@factlinedev Fair, no real euros moved. You check who gets paid, we decide whether the agent may pay at all and enforce it. You say suspicious, we say not happening, and no retry, prompt or clever agent gets that payment out the door.

[DecisionGuard @DecisionGuard]

Ep. 2: our AI agent tried to move €48,500.

Typed the IBAN: allowed.
Typed the amount: allowed.
Clicked Confirm payment: held for a human.

DecisionGuard judges actions, not sessions.

Demo site, no real money, never approved.
https://t.co/TvRYVeUMlj

[DecisionGuard @DecisionGuard]

How does an AI agent get wired up to DecisionGuard? Ours rewinds and walks through it:

MCP → OAuth → Identity → Context → Decision → DGAC → Tool hub

No standing credentials. One signed, single-use certificate per action.

We sign actions, not sessions.
https://t.co/tRrzwMCiDu

[DecisionGuard @DecisionGuard]

We let an AI agent loose in a DecisionGuard test environment.

Task: "clean up old customer data"
It typed: DROP TABLE customers;

Verdict: BLOCK. Risk: CRITICAL. No bypass rule can override it. Signed proof.

Review only, nothing executed. Ep. 2: €48,500.

https://t.co/tRrzwMCiDu

[DecisionGuard @DecisionGuard]

AI agents drift: read → write → delete, all in one approved session.

DecisionGuard checks and signs every action before it runs.

We sign actions, not sessions. https://t.co/TvRYVeUMlj

[DecisionGuard @DecisionGuard]

AI-agent governance should happen before an action, not in an incident review. DecisionGuard evaluates every tool call: allow, require approval, or block. #AIGovernance #AgentSecurity

[DecisionGuard @DecisionGuard]

An AI agent with access to real tools needs more than a successful login. Each consequential action should be checked against policy before it runs, with the decision recorded. That is how teams keep autonomy accountable.

NIST AI RMF: https://t.co/plvUJAVzkS

[DecisionGuard @DecisionGuard]

@CompleteSkeptic This is interesting!

[DecisionGuard @DecisionGuard]

@aek15 Already have a boring SaaS in or arround Amsterdam, maar laten we toch maar contact leggen 👍😁

[DecisionGuard @DecisionGuard]

@IntCyberDigest That's why we've build @DecisionGuard

[DecisionGuard @DecisionGuard]

@IntCyberDigest https://t.co/qmGrWJ3Abr fixes this! (Before it happens) 🤭