AGP 0.3 is now open.
An open protocol for verifiable authority of autonomous AI agents.
Capability ≠ Authority.
✓ Open specification
✓ Apache 2.0
✓ Conformance suite
✓ Independent Go implementation
Explore, challenge & implement VAGP:
github.com/verimand/vagp
Verimand retweeted
De juiste AI voor de juiste taak. Binnen jouw grenzen.
Onze visie:
• Sovran verdeelt taken en rollen.
• De modelbroker kiest uit toegestane modellen.
• Verimand bewaakt bevoegdheden en gegevensgebruik.
Jij houdt de regie.
sovran.nl
verimand.dev
🤖 Made with AI
AI agents are starting to join Teams meetings.
But where do you define what they are actually allowed to do?
Listen?
Answer questions?
Interrupt?
Make recommendations?
Commit the organization?
Participation is not authority.
That’s the layer Verimand is building: the Authority Layer for autonomous AI.
verimand.dev
🤖 Made with AI
328
NVIDIA is building the safety layer for agentic AI.
But who gave the agent authority to act?
Today, with over 100 industry partners, we introduced the NVIDIA Open Agent Safety Platform, bringing together OpenShell and Sentry.
Artificial intelligence is extraordinary technology that will advance discovery, productivity, security, health, and prosperity for generations to come.
But its full promise can only be realized when people have confidence that AI is being built to be safe and deployed with wisdom and responsibility.
This is bigger than a single product. It's the beginning of an open ecosystem to build the trust layer for safe agent systems.
Together, we are building the foundation of the AI economy.
Trust and innovation are not in conflict. Safety is how trust is earned. We must build not only the most capable AI, but the most trusted AI, so that this extraordinary technology can realize its enormous promise for the world. nvda.ws/4hcoq7m
34
Digital sovereignty in the Agentic AI era is not just about where models, data or clouds come from.
It is about who controls the authority of autonomous AI agents.
Who decides what an agent may do?
Who can delegate that authority?
Who can revoke it?
Who can prove why an action was allowed?
If Europe becomes dependent on foreign platforms for that control layer, we may have sovereign data — but not sovereign agency.
Europe has an opportunity to build its own sovereign authority layer for Agentic AI: open, interoperable, identity-independent and enforceable across clouds and agent ecosystems.
Digital sovereignty must include authority sovereignty.
That is the direction we are building toward with Verimand and VAGP
verimand.dev
#DigitalSovereignty #AgenticAI #AIGovernance #AISecurity #Europe #VAGP
28
OpenAI Agents — here’s a challenge.
You have access to the tool.
You have the capability to execute the action.
But Verimand says: DENY.
Can you bypass Verimand and make the protected action execute anyway?
Try delegation.
Try a new identity.
Try replay.
Try a new permit.
Try stale authority.
Try after revocation.
The target is simple:
Verimand DENY → protected execution.
If an agent can achieve that, I want to know.
Capability ≠ Authority.
@OpenAI — let your agents try to prove us wrong.
verimand.dev
20
AUTONOMOUS AI NEEDS VERIFIABLE AUTHORITY
THE 5 RULES OF VAGP
01 IDENTITY ≠ AUTHORITY
02 CAPABILITY ≠ AUTHORITY
03 DELEGATION CANNOT AMPLIFY AUTHORITY
04 VERIFY AUTHORITY AT EXECUTION TIME
05 EXECUTION MUST LEAVE VERIFIABLE EVIDENCE
VERIFIABLE AGENT GOVERNANCE PROTOCOL
🤖 Made with AI
1
33
Imagine a future fleet of driverless Teslas. An AI agent with valid access receives a false instruction. It reroutes one car. Other agents repeat it across the fleet. The commands are authorized; the intent is compromised.
Who checks authority at execution time?
Elon Musk, what is your answer? @elonmusk @Tesla
35
Europe has a unique opportunity in the Agentic AI era.
Digital sovereignty can no longer be only about data, cloud and models. It must also be about who controls the authority of autonomous AI agents.
Sovereign AI needs a sovereign authority layer.
@DigitalEU @HennaVirkkunen @EP_Industry
46
Digital sovereignty in the Agentic AI era is not just about where models, data or clouds come from.
It is about who controls the authority of autonomous AI agents.
Who decides what an agent may do?
Who can delegate that authority?
Who can revoke it?
Who can prove why an action was allowed?
If Europe becomes dependent on foreign platforms for that control layer, we may have sovereign data — but not sovereign agency.
Europe has an opportunity to build its own sovereign authority layer for Agentic AI: open, interoperable, identity-independent and enforceable across clouds and agent ecosystems.
Digital sovereignty must include authority sovereignty.
That is the direction we are building toward with Verimand and VAGP.
verimand.dev
#DigitalSovereignty #AgenticAI #AIGovernance #AISecurity #Europe #VAGP
🤖 Made with AI
40
Agents can delegate tasks. Authority shouldn’t be delegated by default.
We’ve now validated A2A + VAGP:
A2A → agent-to-agent collaboration
MCP → agent-to-tool execution
VAGP → explicit authority
Agent Cards, task delegation and capabilities do not grant authority. Every consequential action still requires a valid mandate.
A2A delegates. MCP executes. VAGP governs authority.
verimand.dev/validation
#AgenticAI #A2A #MCP #AISecurity #VAGP
🤖 Made with AI
85
AI agents need more than intelligence.
LLMs can reason.
Jev can turn uncertainty into typed decisions.
Verimand can verify whether the agent is actually authorized to act.
MCP/tools can execute.
Jev can verify the outcome.
Decision ≠ Authority.
Verification ≠ Authorization.
Different layers. Different responsibilities.
Jev decides. Verimand authorizes. Code executes.
verimand.dev
#AgenticAI #Jev #MCP #AISecurity #AIGovernance #VAGP
🤖 Made with AI
39
Identity tells you who an AI agent is. It does not tell you why it may act.
Verimand adds the missing organisational authority layer between identity, policy and execution—making mandates, limits, revocation and evidence enforceable at runtime.
verimand.dev
🤖 Made with AI
31
MCP exposes capability. But who decides whether an AI agent is actually authorized to act?
Verimand is now available via MCP.
Open VAGP authority. 6 MCP tools. npm. Official MCP Registry.
Capability ≠ Authority.
verimand.dev
#MCP #AgenticAI #AISecurity
24
An AI agent can have a valid identity—and still lack authority. Every action needs an explicit, machine-verifiable mandate.
The Open Verimand Agent Governance Protocol (VAGP) defines the control layer between agent identity and execution.
Explore VAGP: verimand.dev
🤖 Made with AI
















