Digital, Data, AI & Cyber Resilience 2026/27: From tool adoption to a governed operating system

02 / Start with the estate and data map

Start with the estate and data map

Leaders cannot govern what the organisation cannot see. The first control is a current map of critical services, devices, identities, integrations, data flows and external dependencies.

For each critical service, a school or trust needs to know:

• the accountable operational owner and technical owner;
• which pupils, staff or families depend on it;
• what data it receives, creates, shares and retains;
• how identities and privileged access are controlled;
• what other services depend on it;
• where logs and configuration evidence are available;
• the support and end-of-life position;
• how essential work continues if unavailable, and how data is handled at exit.

NCSC's July and August warnings reinforce the need for accurate visibility at the network edge. Its 13 July advisory highlighted secure router and network-device management. On 27 August, it urged organisations to inventory internet-exposed systems, map data flows, patch promptly, retire unsupported devices, remove legacy protocols and exercise incident readiness. These are all-sector threat messages, not evidence that English schools were targeted. Their value is the defensive pattern.

The estate map should therefore identify internet-facing and unsupported technology explicitly.

03 / Create one lifecycle gate for EdTech and AI

Create one lifecycle gate for EdTech and AI

DfE's 9 July guidance provides the strongest basis for a cross-functional approval process. It says schools need data protection by design and default, early DPO involvement, minimum necessary data, clear controller and processor roles, lawful basis, data-flow understanding, security, incident notification, monitoring and exit arrangements. A DPIA is mandatory where processing is likely to result in high risk.

The gate should begin with use rather than brand. A staff drafting tool with no personal data is not equivalent to a pupil-facing chatbot. A system that suggests is not equivalent to one that profiles or decides. An automated agent with credentials and network access creates a different level of exposure from a closed text generator.

A proportionate record should cover purpose, users, benefit and excluded uses; data types and flows; legal roles and lawful basis; DPIA decision; safeguarding and accessibility; moderation and human review; permissions, authentication and logs; incident route; change triggers; and exit or deletion.

The group involved should match the risk. The operational owner defines purpose. Curriculum or inclusion leadership tests educational fit and accessibility. The DSL considers content, contact, conduct and reporting. The DPO examines rights and processing. Digital and IT staff assess integration, identity, logging and recovery. A senior leader accepts residual risk.

Every material service also needs a named owner after approval. A new model, feature, purpose, data type, user group, integration, subprocessor, data location or autonomous action should force reassessment. Without this change control, the school is approving a snapshot while operating a moving system.

ICO's August Children's Code update gives this work an empirical regulatory basis. It reports 2024/25 audits of 28 EdTech providers and 596 recommendations, with recurring weaknesses in contracts, data-flow maps, minimisation and retention, privacy information and DPIAs. Those audits do not measure school compliance and predate this evidence window, but the themes provide a credible internal assurance checklist.

04 / Treat AI as part of safeguarding architecture

Treat AI as part of safeguarding architecture

KCSIE 2026, effective 1 September, makes online safety a running and interrelated theme across policy, curriculum, staff development, DSL work and parent engagement. It includes digitally altered or wholly AI-generated nudes and semi-nudes, generative-AI simulated contact and product-safety considerations. It also links cyber resilience to wider safeguarding.

The 25 August filtering and monitoring update translates that context into an operating trigger. Reviews should occur annually and when safeguarding risk, working practice, major software, configuration or technology such as GenAI changes. The standard recognises real-time, dynamic and personalised content.

This has four consequences.

First, service approval and review of the wider control environment are separate decisions. Existing filtering, monitoring or reporting may remain inadequate.

Second, technical controls are only one layer. ICO research published on 19 August found that 41% of children whose parents used controls had tried to bypass them, and 49% of those children said it was easy. The survey involved more than 4,000 children aged 8 to 17 and parents, with fieldwork from 5 to 30 January 2026. It is not a school-effect study, but it shows why pupil understanding, reporting confidence and practical testing matter alongside configuration.

Third, deepfake response must join safeguarding, behaviour, evidence preservation, pupil support, parent communication and staff training. Schools should not wait for a high-profile incident to decide who leads.

Fourth, the DSL, senior leader, IT lead and responsible governor need a shared review record. A configured product is not proof that the control is effective across all devices, sites, services and working patterns.

05 / Make cyber assurance a board operating rhythm

Make cyber assurance a board operating rhythm

The Academy Trust Handbook 2026 is final and takes effect on 1 October. It says academy trusts must be aware of cybercrime and put proportionate controls and action in place. Trusts must not pay ransoms or extortion demands, including ransomware demands. They should meet DfE cyber standards and work towards six core digital standards by 2030.

The wording matters. The no-ransom and awareness/control provisions are must clauses. Several standards expectations are should clauses. The handbook is a funding-agreement condition for academy trusts, while the maintained-schools governance guide is non-statutory for its audience. The two should not be blended into one universal legal claim.

Both governance routes support a board evidence cycle. It should cover:

• named executive and governor responsibility, with appropriate training;
• current critical-service, internet-facing and end-of-life inventories;
• privileged and remote access, strong authentication and exceptions;
• patching, supported software and secure device management;
• current backups and evidence of successful restoration;
• agreed recovery priorities and service objectives;
• incident exercises involving leadership, safeguarding, communications and operations;
• decision, reporting and escalation routes, including the no-ransom position;
• assurance for externally operated critical services;
• actions, owners, due dates and residual exposure.

The board needs to know whether each control is current, tested and complete, what remains outside it, and how exposure will be handled.

NCSC's 29 July forensic-observability article adds a useful test: could the organisation reconstruct what happened? Logs, telemetry, configuration state and attributable activity matter before an incident, not only after one. NCSC's interim 20 August agentic-AI advice applies the same logic to autonomous systems through restricted authority, monitoring, human oversight and emergency shutdown.

06 / Prepare MIS transition as continuity work

Prepare MIS transition as continuity work

DfE's July MIS guidance aims for a national framework by June 2027. The Academy Trust Handbook says all academy-trust MIS contracts should align by September 2027 and provides transition rules around 1 September 2027.

The core principles place the school or trust's data ownership and controller role at the centre. They cover location and lawful transfers, restrictions on use beyond service delivery, accessible download or APIs, migration-ready data, cooperation at exit, renewal limits and progress towards Cyber Essentials Plus or ISO 27001.

This is not simply a selection exercise. MIS supports attendance, safeguarding, assessment, communication, statutory reporting and daily administration. Transition failure can become an educational and safeguarding continuity failure.

The 2026/27 task is to build a contract-to-data map: expiry and notice dates, decision authority, data dictionary, flows and integrations, access and export routes, security evidence, location, retention, deletion, parallel running, restoration and acceptance tests. A usable export should be tested while there is time to correct gaps.

The planned framework was not live at the cut-off. DfE's 22 July invitation to user research confirms that the wider service is still being designed. Principles may change, and maintained schools are not governed by the Academy Trust Handbook in the same way. Leaders should prepare without claiming a finished national route.

07 / Use regional AI support without outsourcing judgement

Use regional AI support without outsourcing judgement

RISE plans updated on 27 August place AI and EdTech in regional school-improvement support through evidence routes, networks, champions and testbeds. The North West plan includes an AI-supported assessment service.

These are delivery intentions, not validated outcome findings or mandates. Provision is uneven and the North West plan contains dates that appear earlier than the August update, so operational timing needs verification.

Schools can use regional support while retaining their own approval and evaluation thresholds. Every trial should name the problem, comparison, learning measure, staff-workload measure, inclusion and accessibility effects, reliability expectations, privacy and safeguarding controls, and stop or scale criteria. Access to a supported service is not evidence that it improves outcomes.

Ofsted's September inspection information provides a useful boundary. It says AI use is not explicit in the framework, inspectors do not expect schools to use it, and they focus on the quality of decision-making rather than the tool. That AI wording predates the evidence window and should not be reported as an August policy change. It remains a reason not to adopt technology for imagined inspection compliance.

08 / A 90-day operating plan

A 90-day operating plan

September 2026

Implement KCSIE 2026 changes. Complete the online-safety and filtering review, including GenAI and dynamic content. Brief staff and governors on deepfake, simulated-contact and reporting routes. Update privacy notices where processing has changed. Run a cross-functional incident exercise that includes loss of a safeguarding-critical service.

October to November 2026

Academy trusts should apply the handbook's cyber provisions from 1 October. Boards should receive the first evidence pack, including inventories, restore tests, unresolved standards gaps and the no-ransom route. All settings should establish the EdTech and AI lifecycle gate, assign owners to high-impact existing services and record material-change triggers.

December 2026 to March 2027

Reconcile service, identity, data-flow and integration registers. Test exports and restoration for critical services. Map MIS contract and notice dates. Review the first term's incidents, workarounds and overdue actions. Evaluate regional or local trials against predefined pupil, staff, inclusion, privacy and reliability measures.

09 / Board dashboard

Board dashboard

Useful board measures include:

• percentage of critical services with a named owner, current data-flow map and recovery or exit route;
• percentage of internet-facing and unsupported assets accounted for and remediated or under a time-bound exception;
• strong-authentication coverage for privileged and remote access;
• date and result of the last successful restore test for each critical service;
• open high-risk actions from safeguarding, filtering, DPIA, security or service reviews;
• percentage of AI and EdTech use cases with accountable ownership, a DPIA decision, human-review design and change trigger;
• incident detection, containment, recovery and lessons-action closure times, interpreted carefully where numbers are small;
• filtering and online-safety review actions completed;
• MIS export and integration tests completed before notice dates;
• pupil, staff, workload and accessibility outcomes from trials, including adverse effects.

Adoption counts alone say nothing about learning, safety, workload or resilience.

10 / Three scenarios to exercise

Three scenarios to exercise

Base case: controlled adoption

Existing services remain stable, new uses pass through the lifecycle gate, KCSIE and handbook changes are implemented, and MIS preparation proceeds in stages. The test is whether evidence remains current after the initial autumn push.

Stress case: critical service compromise

An internet-facing or externally operated service is compromised during a safeguarding-critical period. Primary systems become unavailable and logs are incomplete. The exercise should test decision authority, containment, alternative processes, communications, recovery priorities, evidence preservation and the no-ransom position.

Change case: service behaviour shifts

A widely used service changes its model, purposes, subprocessors, data location or autonomous capabilities mid-year. The exercise should test who detects the change, who suspends or limits use, which reviews reopen, how staff and families are informed, and what evidence supports continuation or exit.

11 / Dates and watchpoints

Dates and watchpoints

12 / Evidence boundaries

Evidence boundaries

This outlook cannot establish whether school cyber incidents increased or whether cited state-linked campaigns targeted English schools. NCSC evidence is threat context, not incidence data. It cannot claim that RISE AI services improve attainment, workload or inclusion because no in-window outcome evaluation was identified. It does not treat the planned MIS framework as operational or the Cyber Security and Resilience Bill as a current school duty.

ICO survey fieldwork occurred in January 2026, and EdTech audits were conducted in 2024/25. Those reference periods remain distinct from August publication. The ICO says it anticipates a possible EdTech code request, not that one is confirmed. NCSC's agentic-AI advice is interim. The broader Ofsted AI position predates July, while only the August document amendments fall inside the evidence window.

The distinction between legal duties, handbook musts, recommended shoulds, future-effective documents, delivery plans and technical guidance should remain visible in every local paper derived from this outlook.