AI security · Ongoing

Continuous AI red teaming.

A model upgrade, a new prompt or a new tool can quietly undo last quarter’s fixes. AI Secured retests your AI after every change, runs a hands-on red team every quarter and tells you when a new attack trick affects you.

Ask about AI Secured
Retest after every AI changeHands-on red team every quarterMonthly AI safety score
AI Secured · sample● demo
  1. FLAGGED Model upgraded · regression suite started
  2. FOUND Old jailbreak works again on the new model
  3. FIXED Guardrail updated, suite rerun
  4. PASSED Quarterly hands-on test passed
  5. FLAGGED New attack technique added to the suite
In plain words

Continuous AI red teaming is a monthly service that keeps testing your AI as it changes. We turn your first red team into a regression suite of attacks, rerun it after every model, prompt or tool change, add new attack techniques as they appear, and run a hands-on red team every quarter. You get a monthly safety score, and your AI Tested letter stays current.

Coverage
Ongoing, month to month
Includes
Quarterly hands-on red team
Reporting
Monthly AI safety score
Pricing
Monthly fee
Why ongoing

A test is a snapshot. Your AI is a moving target.

One-off red team

A strong snapshot of one version.

  • Deep test of the AI as it is today
  • Out of date after the next model upgrade
  • Misses jailbreaks published next month
  • No warning when a fix quietly breaks

AI Secured (monthly)

The same depth, kept current.

  • Regression attacks after every change
  • New attack techniques added as they appear
  • Hands-on red team every quarter
  • AI Tested letter kept valid
Triggers

What sets off a retest.

  • Model upgrades

    A new model or version from your provider, or a switch to a different one.

  • Prompt changes

    Edits to system prompts, guardrails or personas.

  • New tools

    New actions, plug-ins or MCP servers.

  • New data sources

    New documents or databases connected to AI search.

  • New attack techniques

    Public jailbreaks and research that affect a setup like yours.

  • Incidents

    Anything odd your team or your customers report.

How it works

A rhythm that keeps pace.

  1. Month 1

    Onboard

    Connect to your release process, so we hear about model, prompt and tool changes.

  2. Month 1

    Baseline

    A full red team, which becomes your own regression suite.

  3. Every change

    Watch

    The suite reruns after each change, and we review anything that slips through.

  4. Every 3 months

    Quarterly test

    A hands-on red team with fresh techniques.

  5. Monthly

    Review

    Safety score, trends and advice on a short call.

What you get

Assurance you can show every month.

HANULSECAI Secured monthly report CLIENTSample · insurance claims assistant SUMMARY FOR LEADERS FINDINGS BY SEVERITYCritical0High1Medium3Low4 REGRESSION SUITE · ALL PASSED TREND IMPROVING Sample layout, not client data
  • 01
    Monthly AI safety score

    One number and a trend line leaders can follow, with the detail behind it.

  • 02
    Regression suite

    Your own library of attacks, rerun after every change.

  • 03
    Quarterly red team report

    Hands-on findings with fixes, just like the first test.

  • 04
    Attack alerts

    A short note when a new technique affects you.

  • 05
    AI Tested letter, kept current

    Reissued after each passed quarterly test.

  • 06
    Advice on call

    Ask us before you ship a risky change.

Who it is for

For AI that never stands still.

  • Teams that ship AI changes every week or month
  • Companies whose customers or auditors expect ongoing AI assurance
  • Products where one bad answer would make the news
  • Anyone who has done a first red team and wants the results to stay true

Start with a red team

AI Secured builds on a full first test. If you have not had one, we start with an LLM red team or an agent test, then move to monthly.

Questions

Questions buyers ask before they call us.

Q01Why retest after a model upgrade?
Each model version responds differently to the same instructions. A guardrail that held on one version can fail on the next, and old jailbreaks sometimes start working again. Regression testing catches that before customers do.
Q02What is the AI safety score?
A score from 0 to 100 based on how many regression attacks succeed, weighted by severity, together with the latest quarterly findings. It is HanulSec’s own measure, designed to show direction over time rather than to compare companies.
Q03Does this replace guardrail or monitoring tools?
No. Guardrail products filter traffic in real time. We test whether your whole setup, guardrails included, still holds up. Many teams use both.
Q04How fast do you retest after a change?
We agree the turnaround in the contract, ideally before the change reaches customers, and faster for urgent releases.
Q05How is it priced?
A fixed monthly fee based on the number of AI systems, how often you release changes and the depth of the quarterly test.
Next hops

Related services

Request a quote

Scope it in two minutes.

Tell us what you need protected. A security engineer, not a salesperson, replies with a fixed-price proposal.

Start your quote hello@hanulsec.com
  • Reply within one business day
  • Free 30-minute scoping call
  • Mutual NDA before any detail is shared
  • Fixed price, free retest of fixed findings