June 2026: EBM has filed a UK patent application, working with leading intellectual property firm Boult, to protect the core technology behind our safe-by-design approach to AI agents.
The problem
AI agents are moving from answering questions to taking action: updating records, running code and making decisions in live systems. Most organisations keep them in check with guardrails, which are policies layered on top of broad access. But an agent holding a password can often do far more than the policy intends. That gap is exactly what attacks such as prompt injection exploit. The result: every new agent adds risk, and another layer of costly checks.
Our answer
EBM's patent-pending technology, AxioS, closes that gap. Instead of broad access plus a policy that hopes to hold, agents can only act through tightly scoped ‘Handles’. These are defined in an ontology or knowledge graph built from the organisation's own data and rules. If a Handle has not been granted, the action simply does not exist for that agent. Every action is recorded as it happens. Want to know more? Find out about EBM AxioS.
Why it matters
- Safe by design: control is part of how the agent is built, not a check added afterwards.
- Evidence by default: risk owners and auditors can see what each agent did, and why it was allowed.
- Faster adoption: leaders can deploy AI without adding risk their organisation can't manage.
Filing the patent protects this innovation and our clients' investment in it. For organisations in government and regulated sectors, it is a key enabler: the benefits of AI, without compromising what they have already built.