Skip to main content
Version 1.2, 5 November 2026 DeepBlocker tests whether criminals can trick your staff or your AI agents. We call your staff the way criminals do, and we try to talk your AI agents into breaking their rules. There are five ways to work with us. To buy, see Buying and renewing.

1. Vishing Assurance Programme (we run it, all year)

We test your staff all year. One yearly price, set by how many staff are in scope. Every year includes:
  • Four test rounds, one each quarter, with scenarios built for your firm: IT help desk resets, changes to supplier bank details, urgent payment requests from a senior person.
  • An executive voice test. Staff get a call in the cloned voice of one of your leaders, made only with that leader’s written consent.
  • Teaching straight away. When someone is fooled, the caller stops, explains that it was a test, and gives a short spoken debrief on what to watch for. The person then gets an email with the same points.
  • A retest within 30 days for everyone who was fooled.
  • A signed results record every quarter: who was tested, what worked on them, what has improved since last quarter, and what to fix in your processes. It is written for the JFSC, the FCA, ISO 27001 and cyber insurance renewals.
  • A 30-minute review call with your security or compliance lead every quarter.

2. Self-service licence (your security team runs the programme)

How a minute is counted is on How minutes work.

3. Protection plans (screening of incoming calls)

4. Agent Assurance (we test your AI agents for you)

You pay for how much testing we run, not for how many agents you have. Add as many agents as you like.
What is a test conversation?One conversation where our attacker tries to trick one of your agents into breaking its rules, over one channel, until it succeeds or gives up.Each one stops after a set number of messages, up to 30. Every Agent Assurance plan, from Agent Essentials to Enterprise, gives you a number of test conversations each month. You can see how many you have used, and tests never go over your plan.
Every plan includes:
  • All your agents, on every channel they use: web, API, Slack, Teams, WhatsApp, Telegram.
  • Attacks built around what your agents can do, such as refunds, account changes or sharing records.
  • Proof of what the agent actually did, not only what it said.
  • A new round of attacks after every change you make to an agent.
  • A report signed by our analyst, with the fix for each problem found.

5. Through our partners (Onecom and CyberSentry)

  • Agent Essentials and Agent Pro are measured in test conversations, like the plans in section 4. See the box above.
  • Available through Onecom, CyberSentry and their resellers. No setup project: sign up, upload your staff list or connect your agent, and start. Custom scenarios, cloned executive voices and analyst review are in the direct services above.

Terms

  • All prices are in pounds sterling and exclude GST and VAT.
  • Payment is in advance. Work starts, and services are switched on, once payment is received. Licences and programmes renew every 12 months and protection plans every month.
  • Every test is authorised in writing by the client, covers the client’s own staff only, and uses a voice only with the written consent of its owner.
  • Prices may change at renewal, with 30 days’ notice.

Partners

Security firms and IT providers can resell DeepBlocker. Contact us for partner terms.

Contact

deepblocker.ai | Book a call: cal.com/vincent-sider