By Micky Irons, founder and CEO of Mickai.
A telecoms operator sees almost everything. It sees where a subscriber is, who they call, which cells are congesting, which routes are failing over, and which fraud pattern is forming three hops upstream before a single complaint lands. No other institution holds a behavioural map of a nation this detailed, this fresh, or this continuous. It is the closest thing to a live nervous system a country has.
The uncomfortable truth is that most operators cannot fully use that map. The moment they reach for modern intelligence to read it, the prevailing pattern asks them to move network telemetry and subscriber records to a general-purpose cloud, trust a control plane they do not own, and accept that the audit trail lives somewhere they cannot independently verify. For a business governed by lawful-intercept duties, retention rules, and national-security expectations, that is not a technical detail. It is the whole problem.
The data an operator cannot afford to export
Call detail records, location traces, deep packet inspection metadata, SIM provisioning logs, and roaming exchanges are among the most sensitive data any commercial organisation holds. Under the General Data Protection Regulation (GDPR) they carry special obligations, and in many jurisdictions telecoms metadata attracts its own retention and disclosure regime on top. A single careless egress is not a bug ticket. It is a regulatory event with named accountable officers.
The instinct to centralise this data into a hyperscale platform is understandable, because that is where the fastest intelligence has historically lived. But an operator that exports its subscriber graph to run analytics on it has already conceded the one thing regulators, and increasingly subscribers themselves, care about most: that the data stays inside a boundary the operator controls and can prove it controls. The answer is not to weaken the intelligence. It is to move the intelligence to the data.
A sovereign intelligence layer inside the operator
Mickai is a Sovereign Intelligence Operating System, a SIOS. It is built and live, and it is designed to run on hardware the operator already owns, air-gapped or on-premise, inside the core network's own trust zone. Its brains reason over network and subscriber data where that data already sits, so nothing has to cross a perimeter to become useful. There is zero data egress by design, not by policy promise.
Like Argus of the hundred eyes, an operator sees everything, and the only real question is who is allowed to look
This is a different layer from the public cloud, not a competitor to it. The hyperscalers remain allies for everything that belongs in shared infrastructure. Mickai serves the regulated boundary that a general-purpose cloud, by its nature, cannot cross on the customer's own terms. An operator can keep its Operations Support Systems (OSS) and Business Support Systems (BSS), its data lakes, and its cloud partnerships exactly as they are, and still hold the intelligence layer inside its own walls.
Every action signed before it runs
Automation in a live network is only as trustworthy as its accountability. When a brain proposes to reroute traffic, throttle a congesting cell, quarantine a fraudulent range, or flag a subscriber for investigation, the operator needs to know exactly what was decided, on what evidence, and by whose authority. It needs that record to be impossible to forge after the fact.
Each action is forged into a hash-linked chain, and a broken link is proof that someone tried to tamper
In Mickai every action produces an Operation Attestation Record (OAR) that is signed before the action executes, not after. The OAR captures the decision, the inputs, and the approving authority, then commits to a hash-linked chain built on SHA-3-512, so any later tampering breaks the chain and shows. Signatures use post-quantum cryptography, specifically the FIPS 204 ML-DSA-65 standard, so the ledger stays defensible against the harvest-now-decrypt-later threat that matters enormously to infrastructure meant to last decades.
An audit trail regulators can verify without trusting the vendor
The strongest property of a signed ledger is that it does not require faith. A telecoms regulator, an internal auditor, or a court-appointed examiner can take the tamper-evident, cryptographically-signed audit ledger and verify it offline, against the operator's own keys, with no call home to Mickai and no dependency on any cloud being reachable. The proof travels with the evidence.
That reframes the compliance conversation. Instead of an operator asserting that its automated systems behaved correctly, it can hand over a verifiable record that demonstrates it. Under the Digital Operational Resilience Act (DORA), the second Network and Information Security Directive (NIS2), and the emerging governance expectations of the EU Artificial Intelligence Act (EU AI Act), that shift from assertion to proof is quickly becoming the difference between a passed audit and a finding.
High-stakes actions need more than one signature
Not every decision should be autonomous. Suspending a subscriber, opening a lawful-intercept workflow, or making a change that touches national infrastructure demands human judgement and layered authority. Mickai supports multi-brain plus voice-biometric approval for high-stakes actions, so a sensitive operation can require independent brains to concur and a named human to authorise with a biometric signature that itself lands in the ledger.
The ledger asks for no faith, only that the scales balance whenever anyone chooses to check them
Brains are also revocable. If a capability is compromised, deprecated, or simply no longer trusted for a given task, it can be revoked cleanly, and every action it ever took remains attributable in the chain. That combination, granular authority plus permanent attribution, is what lets an operator automate aggressively without ever losing the thread of who did what.
What this unlocks across the network
With the intelligence inside the boundary, the use cases that data-residency fear once blocked become ordinary. Real-time fraud and SIM-swap detection can read the full signalling picture. Predictive maintenance can reason over raw radio telemetry that was too sensitive to export. Capacity planning, roaming settlement anomaly detection, and churn intelligence all draw on the complete subscriber graph rather than a thinned, tokenised shadow of it.
The most sensitive gates open only when several guardians agree and a named authority speaks
The intellectual property behind this is deliberate. Our approach is captured across 104 filed UK patent applications, about 2,340 claims, owned by Mickai LTD, and framed around the capabilities that matter here: signing before execution, offline verification, hardware-bound sovereignty, and layered biometric authority. The point of the patents is not the count. It is that the sovereign-boundary architecture an operator needs is a designed, defended capability rather than an afterthought.
The bottom line
Telecoms operators have been asked to choose between powerful intelligence and provable control, and for years the honest answer was that they could not have both. Mickai removes the trade-off. The intelligence comes to the data, every action is signed before it runs, and the whole record can be verified offline by anyone the operator has to answer to.
For an industry that carries a nation's private conversations and its critical connectivity, sovereignty is not a premium feature. It is the baseline. An operator that keeps its network and subscriber-data intelligence inside its own signed, auditable boundary is not just safer. It is finally free to use the extraordinary dataset it has always held.
Written by Micky Irons. Originally published at https://mickai.co.uk/articles/sovereign-ai-for-telecoms. More from Micky Irons and Mickai at mickai.co.uk.
Top comments (0)