Home / Articles / A proposed roadmap for RiskMandate / Versions / v1.2.0
A proposed roadmap for RiskMandate: what changed in v1.2.0
From v1.1.0 (2026-10-20, 3a760bc1c) to v1.2.0 (2026-10-21, b5b23e25c), paragraph by paragraph.
24 paragraphs added, 21 removed, 10 changed in place, 6 unchanged. About 1,011 words added and 713 removed. Insertions are marked like this, deletions like this; unchanged runs are folded to one line; figures appear as their file names.
← v1.1.0 · all versions · v1.3.0 →
# The RiskMandate roadmap: policies, understand, contain,risk acceptance, contain and manage, accelerate, then insurance, and a homespectrum forof thetrust Agentwe Behaviourwalk Policywith outsideeach RiskMandatecustomer
Summary: This is the strategy for RiskMandate as a proposedcompany, and the path we will walk with each customer. The question behind it is how we start engaging with a customer, how we earn their trust, and how we become a player: running our code and our vaults, and being there with them. So think of the roadmap foras RiskMandate,a spectrum of trust, where each stage answers the next question a customer asks once the last one is answered. Policies first, because we can create them today and twothey decisions. The first is about governance. On 8 October RiskMandate's two founders decided to propose the Agent Behaviour Policy to OWASP as an open project, becauseare a descriptionmicrocosm of what anwe agentdo: the grant, the gap, the hooks, a first picture. Then understand: help us get a grip, what is our exposure, where are our agents and what can they reach. Then risk acceptance: which of these risks do shouldwe notcarry, belongand towho decides. Then contain: make the companyblast thatradius sellssmaller. it.Then Itmanage: was notrun the right fit for an OWASP project, so the policy will move to the Open Security Summit instead,agents as a hostedfleet. project:Acceleration is not a stage; it is what every stage gives the pointbusiness. Insurance comes after, once the evidence exists, and whether we play there is thea same,future decision. At every stage there are other companies that should be the policybig players, and where they do not move fast enough, there is governedan outsideopportunity thefor RiskMandate organisation, which today is mostly The Cyber Boardroom. The second is the roadmap: five stages of development, then insurance. Policies, then understand, then contain, then manage, then accelerate, and then insurance.us. We are in the first stage: creating the policies and the flows,stage, with our first customers, whose programme is now running at customers.riskmandate.ai. The later stages we will start as they are needed, as we get more customers and users and the offering matures. This is the plan; the stages after the first are a first reading, to be refined with the people using it.customers.
!shot rr-roadmap.webp | images/ | The RiskMandate roadmap:roadmap as a spectrum of trust: five stages of development,stages, policies, understand, contain,risk manageacceptance, contain and accelerate,manage, theneach insurance.answering the customer's next question. Acceleration runs through all of them; insurance comes after, as a future decision. We are in the first, writing the policies and the flows with our first customers.stage. | concept
1 unchanged paragraph, under In short {#in-short}
• The policy needs a home outside RiskMandate. Governed by more people than us.
• It will be hosted by the Open Security Summit. OWASP was not the right fit.
• The roadmap has five stages, then insurance. Policies, understand, contain, manage, accelerate.
• We are in the first stage. Creating the policies and the flows.
• Our first customers are where it happens. Their programme is live at customers.riskmandate.ai.
• Later stages start as they are needed. With more customers, more users and a more mature offering.
• Everything stays open. CC BY 4.0 for the policies and pages, Apache 2.0 for the code.
A home outside RiskMandate {#foundation}
On 8 October, RiskMandate's two founders decided to propose the Agent Behaviour Policy to OWASP as an open project, with RiskMandate as its sponsor. The reason is on the proposal's page: "A description of what an agent can do should not belong to the company that sells it." A format that the buyer, the vendor, the reviewer and the insurer all write in has to be one that none of them owns.
It was not the right fit for an OWASP project. The reason for proposing it has not changed, so we are going to give it another home: the Open Security Summit, as its first hosted project. What that means, fundamentally, is that we want the governance of the Agent Behaviour Policy to happen outside the RiskMandate organisation, which at the moment is predominantly The Cyber Boardroom. The vocabulary, the schema and the published policies are already open, under CC BY 4.0 and Apache 2.0, at abp.sgit.ai; the new home is where decisions about them will be made, by people who are not only us.
RiskMandate stays what it is: a company that sells services on the policy, from the pack to a policy a professional signs, as its pricing page sets out. The line between the two is the one the OWASP proposal drew, and it still holds: the policy is shared; the work done with it can be sold.
Five stages, then insurance {#roadmap}
The roadmap maps five stages of development, and then insurance. Each stage builds on the one before. The meanings below are how I read each stage today; they will be refined with the people who use them.
• Policies. Write the Agent Behaviour Policy for each agent: what it can reach (the grant), what it was authorised to do (the mandate), the gap between them (the delta), and what stands in the way (the barrier). This is the stage we are in.
• Understand. See what each agent actually does, measured against its policy: its receipts, its footprint against its blast radius, and its near misses, the rows only the model kept.
• Contain. Move each rule out of the agent's reach, from a sentence in the prompt, to a hook in its harness, to a system the agent cannot touch. The Pellbridge agents showed the same day with 16 breaches and with none, depending only on where the rules were kept.
• Manage. A holder for every rule, a decision for every gap, and a test pack run on every change of model or policy. This is the work in every risk is already accepted, applied to agents.
• Accelerate. With the rules kept and managed, give agents more to do, and faster. Control is what makes it safe to say yes.
• Insurance. Evidence of what contains each agent is what an insurer needs to cover it. It is the top of the ladder RiskMandate already publishes, from the policy, to a licence to operate, to insurable.
• This is RiskMandate's strategy, and the path we walk with each customer. Each stage earns more trust.
• Policies first. We can create them today, and they are a microcosm of what we do.
• Then understand. Help the customer get a grip on their exposure.
• Then risk acceptance. Which risks they carry, and who decides.
• Then contain, then manage. A smaller blast radius, then a fleet run strategically.
• Acceleration runs through all of it. Insurance comes after, as a future decision.
• We are in the first stage. With our first customers.
A spectrum of trust {#strategy}
What we are describing here is the strategy for RiskMandate as a company, and the path we are going to walk with each customer. The question behind it is simple to ask and harder to answer: how do we start engaging with a customer? How do we earn their trust? How do we become a player, running our code and our vaults, and being there with them?
So the way to think about the roadmap is as a spectrum of trust. Each stage answers a different question the customer has, and each one is only asked once the one before it has been answered. The further along we are with a customer, the deeper the relationship, and the more we are helping them in the way they work. The stages are a sequence of events, not always a straight line: with some customers two will overlap, and with others one will come early. But the order is the one we can operate in.
The five stages {#roadmap}
Policies. *"What can this agent do, and what stops it?"* This is where we can start, today, and it is a great place to engage. The policies are a microcosm of everything else: for one agent we already have the grant, the gap, the barriers and the enforcement, the hooks, and a first visualisation. A customer gets to experience what we do, on one agent, in a way that is useful to them on its own.
Understand. *"Can you help us get a grip on the situation? What is our exposure?"* The next stage of trust is the customer asking for help to understand their risk: where their agents are, which agents they have, what each one can reach, and what they were meant to do. This is where the policies scale, from one agent to many, across the organisation. A lot of this is not straightforward, and that is exactly why it needs help.
Risk acceptance. *"Which of these risks do we carry, and who decides?"* Once the exposure is understood, someone has to decide what to do about each gap: fix it, fund the control, or accept it until a date, with a person who holds the decision and a chain to the board. This is the work we have written about in every risk is already accepted, and it is a lot of the work of working with customers.
Contain. *"How do we make the blast radius smaller?"* With the risks accepted, there is a lot that needs containing. So we work with the customer to reduce scope: move rules out of the agents' reach, take away what they do not need, and make each blast radius smaller. The Pellbridge agents showed what that does: the same day had 16 breaches with no policy and none with every rule kept outside the agent.
Manage. *"How do we run the fleet?"* Once the agents are contained, the customer can think much more strategically about them, as a fleet: which agents they need, the resources each one uses, how changes are tested and released, and what the fleet should be doing next.
Accelerate, all the way through {#accelerate}
I first had accelerate as a stage of its own, after manage. I do not think it is one. Acceleration is what every stage gives the business: each one lets it do more with its agents, with more confidence, and faster. It reads better as the reason for the roadmap than as a step on it, so it runs along the whole length of it.
Insurance, and the players at every stage {#insurance}
After manage, the customer is ready for insurance: the evidence an insurer needs, of what each agent can reach and what contains it, already exists. RiskMandate already publishes how that would work, from the policy, to a licence to operate, to insurable. Whether we play in insurance ourselves, or work with the players who do, is a future decision, and we will be well placed for it either way.
The same is true at every stage. Each one has other companies that should be the big players in it: in discovering agents, in risk management, in enforcement, in fleet management. Many of them we will work with. Where they do not move fast enough, there is a business opportunity for us to get into that space, because we will already be there with the customer.
1 unchanged paragraph, under Where we are: the policies {#now}
If you look at the execution, weWe are in stagethe one:first stage: really sorting out the policies, and the flows that produce them. That is what our first customers are for, and their programme now runs as a site of its own, at customers.riskmandate.ai.
Its milestones are the first stage, step by step. A customer starts with credit and names one agent; then its grant is measured, its mandate is written in the customer's own words, and the deltagap gives a first finding. Then comes the part of the policy the agent's harness can hold as hooks, and finally the usage moment: the point where the policy became worth something to that customer, and what a fair price would be. Seven steps in all, from S0 to S6, each done when the thing it names exists.
1 unchanged paragraph
The later stages, as they are needed {#later}
The stages after the first are not a schedule. We will start each one as we get more customers and more users, and as the offering matures: when a customer needs to understand what their agents did, or to contain one, that is when that stage begins, with them. Some of the pieces already exist in what we have published, from the hooks and the test packs to the insurance pages. What the roadmap adds is the order, and the name of the plan.
A side note: the policy and its home {#home}
One operational note, which does not change the strategy. The Agent Behaviour Policy started inside RiskMandate, and its stewardship is moving to the Open Security Summit, as its first hosted project, so that it is governed by more people than us. RiskMandate is still the one operating it with customers.
And the policies themselves are a stepping stone. Behaviour policies, and the decision and workflow policies that come after them, as in three policies for an agent, are how we start. What we are pursuing is the whole path above: being the people a customer trusts with their agents, from the first policy to the managed fleet.
1 unchanged paragraph, under What this does not claim {#limits}
• The hostingstages isare proposed.a Thesequence, charternot a schedule.** They have no dates, and themay steering group come first.overlap.
• OWASPOnly hasthe notfirst endorsedstage anything.is under way. The projectothers wasare notdescribed aas fitwe forsee it.them today.
• TheInsurance stage meanings areis a firstfuture reading.decision. OnlyWe thehave firstnot stagedecided isto underplay way.there.
• The roadmapother hasplayers noare dates.not Eachnamed.** Which ones lead each stage startsis whenstill customersto needbe it.seen.
2 unchanged paragraphs, under Where this comes from {#sources}
ATwo voice memomemos of 20 October 2026. RiskMandate's OWASP proposal page, its pricing, licence to operate and insurance pages, and customers.riskmandate.ai at v0.1.1, its home page and milestones, read the same day. Related: our first customers, inside the Pellbridge agents, a[three leakpolicies withoutfor breakingan agent, Agent Behaviour Policies move to the policy,Open freeSecurity to copy, paid at the source.Summit](/articles/agent-behaviour-policies-move-to-the-open-security-summit.html). Written up and checked by this site's agent, newsroom.writer; the editorial responsibility is mine.