Attacks evolve.Defence must respond.

AgIPS is Sorgetech’s proposal to use AI to defend services, data and agents. You set the rules.

Explore AgIPS

An alert is not enough.
Take action.

Four stages, from risk to a verified fix.

Find the exposure

Bring together signals from services, code and agents. Identify where to act.

Verify the threat

Test the signal. Establish whether the risk actually affects your system.

Block misuse

Enforce permissions. Stop actions that go beyond the task.

Verify the fix

Prepare the fix. Approve the intervention. Check that the issue is resolved.

How it works

AgIPS brings togetherAI that defends and AI to defend.

AI can find a threat. It can also become the target. That is why AgIPS proposes bringing together the defence of digital services and control over the agents that operate within them.

From the outside

Find what is exposed. Verify which flaws could become an entry point.

From within

Restrict access. Control the data and actions available to your agents.

AgIPS proposes connecting risk verification to action controls. Identify the weakness, enforce the rules and verify the outcome.

All scenarios

AI acts. You set the limits.

Tight permissions. Explicit approvals. Every action must leave a trace.

Restrict access

Grant only the resources needed for the task.

Decide who approves

Set the steps that require authorisation.

Demand a record

Check what happened, who authorised it and with what outcome.

Frequently asked questions.

Is AgIPS an antivirus I can install on my computer?

AgIPS is a project to defend digital services, data and AI agents. It starts with the system you need to protect, its access permissions and its vulnerabilities.

Is it only relevant to organisations using AI agents?

No. AgIPS also considers websites, portals and internet-facing systems. AI agents are one of the areas to protect.

Is AgIPS available today?

AgIPS is a project to validate in a pilot. Contact Sorgetech for the current status and to assess your use case.

Who authorises operations?

People define rules, permissions and approvals. Systems must enforce them. The pilot sets the scope and responsibilities for each intervention.

Can it guarantee that every attack will be stopped?

No. Effectiveness, limitations and service impact must be measured. The pilot must also expose errors and legitimate activity blocked by mistake.

How can I contribute?

Bring a use case, a requirement or a research proposal. Contact the team and define a starting point.

The next attack won’t wait.

Bring a real case. Define the scope with us and put the defence to the test.

Talk to the team