The Same Machine
By Amul Patel
2026-09-18T00:00:00.000Z
A briefing for schools, multi-academy trusts, charities and small businesses on what the convergence of information manipulation and financial crime means in practice, what the 2026 evidence shows, and what can be done about it in ninety days without a budget.
A briefing for schools, multi-academy trusts, charities and small businesses on what the convergence of information manipulation and financial crime means in practice, what the 2026 evidence shows, and what can be done about it in ninety days without a budget. Key figures at a glance Every figure in this briefing comes from a named primary source. Where two sources point in different directions, that is stated rather than resolved in favour of the more striking number. Figure What it measures Source 80%+ Share of global phishing campaigns using AI-generated or AI-enhanced content ENISA Threat Landscape 2025 540 Foreign information manipulation incidents analysed in 2025 EEAS 4th FIMI Report, March 2026 £1.28bn UK payment fraud losses in 2025 UK Finance Annual Fraud Report 2026 66% Share of UK APP fraud cases originating online UK Finance Annual Fraud Report 2026 5.19m Estimated cyber crimes against UK businesses in 12 months DSIT / Home Office CSBS 2025/2026 96% Breached secondary schools where phishing was involved CSBS 2025/2026, Education Annex 79% FE and HE institutions experiencing impersonation of the organisation or its staff CSBS 2025/2026, Education Annex 17% UK charities running any staff cyber training, down from 21% DSIT / Home Office CSBS 2025/2026 1. Two stories, one supply chain The security world tells these as separate stories. The first is geopolitical: foreign states, influence operations, coordinated inauthentic behaviour, elections under pressure. It is discussed in select committees and think tanks. It feels like somebody else’s problem if you run a primary school in Buckinghamshire or a four-person charity in South London. The second is criminal: phishing, fraud, business email compromise, ransomware. That one is obviously your problem, and you have been told so repeatedly this year. The 2025–26 evidence says these are not two industries. They are one industry with two customer bases. The accounts are the same accounts. The synthetic content comes from the same tools. Distribution runs through the same platforms and the same recommendation algorithms. The psychological mechanism is identical: move the target emotionally before they engage analytically. What differs is the payoff. The propagandist wants belief. The criminal wants a payment. Everything upstream is shared. That matters to under-resourced organisations for a reason unrelated to politics. When two threat economies share a supply chain, capability improvements in one immediately benefit the other. Better synthetic personas built for influence work make better fake suppliers. Cheaper voice cloning built for disinformation makes cheaper fraudulent calls to your finance officer. You inherit capability developed against an adversary you will never meet. 2. The scale of the machine 2.1 The information manipulation side The European External Action Service detected and analysed 540 foreign information manipulation and interference incidents between 1 January and 31 December 2025, attributing 29% to Russia and 6% to China, and noting that the two infrastructures occasionally act in concert through opportunistic amplification of each other’s content. The World Economic Forum’s Global Risks Report 2026 again placed mis- and disinformation among the top short-term global risks, alongside geoeconomic confrontation and societal polarisation. The point worth carrying forward is not the attribution. It is that this is now a measured, industrialised activity with documented infrastructure, budgets and tooling. 2.2 The criminal side UK Finance’s Annual Fraud Report 2026 sets out the other half. Metric 2025 figure Change on 2024 Total payment fraud losses £1.28bn +4% Confirmed fraud cases 4.06 million +11% (+31% since 2023) Authorised push payment (APP) losses £576.4m +19% APP fraud cases 248,070 — Investment fraud losses £221.5m +40% Unauthorised fraud losses £703.4m −5% Reimbursed to APP victims £354.3m 61% of losses On average, eight people were defrauded every minute in the UK in 2025, losing almost £2,500 between them in that time. The origin point is unambiguous: 66% of APP fraud cases originated online, with a further 17% beginning through telecommunications. UK Finance specifically identifies investment fraud as frequently starting through fraudulent advertising on social media, and purchase fraud as frequently starting on marketplace sites. Its managing director of economic crime has said publicly that the banking sector cannot be the only line of defence. The asymmetry that matters Unauthorised fraud — where criminals attack systems — is falling. Authorised fraud, where criminals persuade a human being to move the money themselves, is rising sharply. The attack surface has migrated from the network to the person, and the person is reachable through social platforms. 2.3 The shared tooling ENISA analysed 4,875 incidents across the EU over a twelve-month reporting period. Two findings connect the halves. Phishing and its variants — vishing, malspam, malvertising — accounted for roughly 60% of all intrusion attempts, and ENISA estimates that more than 80% of global phishing campaigns now use AI-generated or AI-enhanced content. The report also documents convergence directly: the lines between hacktivists, cyber criminals and state-sponsored groups are blurring, with shared tools, tactics and infrastructure increasingly common. It records “faketivism”, where state-aligned intrusion sets operate behind hacktivist personas, and cases such as FunkSec blending political messaging with financial extortion, showing how quickly ideology-driven branding pivots to monetisation. The AI providers observe the same convergence from the supply side. OpenAI’s February 2026 threat report covers emotionally manipulative dating scams, impersonation of legal professionals, and state-backed influence operations in a single document. It describes threat activity that rarely stays on one platform: actors combine AI with social media accounts and fake websites, switching between models for different tasks. Since public threat reporting began in February 2024, more than 40 networks violating usage policies have been disrupted and reported. Different objectives. Same workflow. Same report. 3. Where this lands: the UK evidence The Cyber Security Breaches Survey 2025/2026, commissioned by DSIT and the Home Office and carried out by Ipsos between August and December 2025, surveyed 2,112 businesses, 1,085 charities and 577 education institutions. It is the most authoritative UK measurement available, and it is blunt. 3.1 Prevalence 43% of businesses and 28% of charities identified a cyber security breach or attack in the last 12 months — approximately 612,000 UK businesses and 57,000 UK charities. Phishing remained the most prevalent type of breach or attack by a wide margin, experienced by 38% of businesses and 25% of charities, and was named the most disruptive type of attack by 69% of those affected. The concentration is worth pausing on. Among organisations that experienced any breach, the proportion experiencing phishing and nothing else has risen — from 45% to 51% for businesses and from 46% to 57% for charities. Attackers are not diversifying. They are doubling down on the vector that works. 3.2 Cyber crime and repeat victimisation 19% of businesses and 14% of charities were victims of at least one cyber crime in the past year — approximately 267,000 businesses and 28,000 charities. Phishing cyber crime dominated, accounting for 93% of businesses and 95% of charities that experienced a cyber crime. Scaled up, UK businesses experienced approximately 5.19 million cyber crimes of all types in twelve months, and UK charities approximately 525,000. The median number per victim organisation was three, but the mean was 19, indicating a high level of repeat victimisation among some organisations. That gap between median and mean is the most operationally useful number in the survey. Most victims are hit a few times. A minority are hit constantly. If you are in the second group, you are being worked systematically, not randomly. 3.3 Cyber-facilitated fraud An estimated 3% of businesses and 1% of charities were victims of fraud resulting from a cyber breach — approximately 43,000 businesses and 3,000 charities, with an estimated 130,000 cyber-facilitated fraud events across the UK business population. Among those incurring a cost, the median was £500, with most falling between £150 and £5,000, and the top 10% of cases reaching £15,000. For a charity with a £120,000 annual income, a £15,000 event is not an IT incident. It is a programme. 4. Education: where impersonation has already arrived The education annex is where the convergence thesis stops being theoretical. These figures cover UK state schools, colleges and universities. 4.1 Prevalence by tier Institution type Identified a breach or attack Attacked at least weekly Primary schools 49% 14% Secondary schools 73% (up from 60%) 20% Further education colleges 88% 24% Higher education institutions 98% 29% Businesses (for comparison) 43% — The rise among secondary schools, from 60% to 73%, was statistically significant. For around one in ten further and higher education institutions (12%) and secondary schools (11%), attacks were being experienced at least daily. 4.2 What kind of attack Attack type Primary Secondary FE / HE Businesses Phishing 90% 96% 96% 88% Impersonation of the organisation or its staff 31% 44% 79% 28% Takeover of website, social or email accounts 2% 6% 18% 5% Viruses, spyware or malware 11% 15% 51% 16% Denial of service 3% 7% 49% 6% Ransomware 0% 6% 14% 3% Base: institutions that identified a breach or attack in the last 12 months. The impersonation line is the story. Impersonation attempts against further and higher education institutions rose from 68% to 79% in a single year, with increases also recorded for malware (42% to 51%), denial of service (36% to 49%), and unauthorised access by staff (22% to 29%) and by students (11% to 23%). Universities are the canary. They are open, publicly documented, populated by tens of thousands of people whose names and roles are published, and therefore among the easiest organisations in the country to impersonate convincingly. Schools and charities share the same structural characteristics at smaller scale. 4.3 An honest discrepancy The same survey records impersonation falling among businesses and charities. Business impersonation breaches decreased to 12%, down from 17% in 2023/2024, and among charities fell to 7%, down from 11% the previous year and 12% two years ago. Two readings are available, and the survey does not settle between them. Either impersonation attempts against smaller organisations genuinely declined, or they became good enough that fewer organisations identified them. The survey itself carries the caveat that it only includes breaches organisations were able to identify and willing to report, and that hidden and unidentified attacks mean the findings may underestimate the true picture. Given that the same period saw an eleven-point rise in impersonation against the best-resourced tier of the sector, the second reading deserves weight. Detection capability and attack volume are not the same variable, and organisations with no monitoring cannot report what they never saw. 4.4 The AI exposure AI tools had already been adopted by 82% of further education colleges, 63% of higher education institutions and 53% of both primary and secondary schools — against 21% of private sector businesses. But only 56% of primary schools, 59% of secondary schools, 66% of further education colleges and 49% of universities that used or were considering AI had specific cyber security practices in place to manage the associated risks. Education has adopted AI faster than business and governed it more slowly than its own adoption rate. That gap is the exposure. 5. The 17% problem Here is the finding that should concern anyone working with charities. Only 19% of businesses and 17% of charities ran any staff training or awareness raising on cyber security in the last 12 months. The charity figure has declined from 21%, driven by low-income charities, where it fell from 18% to 13%. Cyber security as a high priority for charity leadership fell from 68% to 60%, its lowest level since 2017/18, driven by charities with income under £100,000, where it dropped from 64% to 53%. Charities seeking any external guidance also fell; among low-income charities it dropped from 32% to 21%. So in the year the attacker’s cost of a convincing, personalised, AI-written approach fell to near zero, the smallest UK charities reduced their attention to it. That is not negligence. It is the arithmetic of an organisation choosing between a beneficiary and a training session. The rest of the control picture matches. Control Businesses Charities Two-factor authentication 47% 38% Agreed process for reporting a fraudulent email 58% 36% Formal incident response plan 25% 19% Board or trustee responsibility for cyber security 31% 30% Reviewed immediate supplier cyber risk 15% 9% Technical controls in all five Cyber Essentials areas 24% 13% Cyber Essentials certified 5% 3% Two-factor authentication — the single control that most reliably defeats credential phishing — is absent in more than half of UK businesses and nearly two-thirds of charities. 6. The detail that proves the argument Buried in the education annex is a school IT lead describing their own phishing simulation. They actively phish their own staff. Click rates had been falling. Then they ran a very convincing email about payroll changes, and the number taken in jumped from 5% to 17% across the trust. They used the data to drive training for those who had done poorly. Read that again. Same staff. Same training. Same institution. The only variable was the quality and relevance of the lure, and it tripled the failure rate. Everything in section 2 is about the cost of producing that lure collapsing. AI writes the payroll email. Social media supplies the names, the roles, the timing and the trust signals. The state-aligned operator and the fraudster draw from the same well. A 5% click rate is manageable. A 17% click rate against an organisation with no MFA, no incident plan and no named owner is a breach with a date on it. 7. The regulatory picture Change is coming, and it is worth knowing where it stands. Ofcom published a package of draft materials on 10 July 2026 as the third phase of Online Safety Act implementation, comprising the register of categorised services and additional duties for the UK’s largest services. Alongside it, two consultations opened: one covering Category 1 duties on terms of service, complaints, user empowerment and identity verification, and a separate consultation for Category 1 and 2A services on fraudulent advertising duties. Both are open until the end of September 2026. UK Finance is pressing for Ofcom to place stronger and proactive fraud prevention obligations on high-risk platforms, and for online marketplaces to verify sellers and use secure payment mechanisms rather than allowing off-platform bank transfers. It also wants a financial contribution from technology and telecoms firms towards prevention. A separate Payments Association white paper, drawing on interviews with Barclays, Revolut, Nationwide and Santander UK, argues that banks bear the cost of scams that begin on social platforms, and that a regulatory imbalance exists because reimbursement duties apply to payment firms while the platforms where fraud originates face no equivalent financial liability. The direction of travel is that platforms will carry more duty. The honest reading of the timetable is that this is consultation now and implementation later. It does not change your exposure this term. 8. What to do, and what the data says about each Every control below is free or close to it, and every one is justified by a specific figure above. Out-of-band verification as a written rule. Any change to bank details, any urgent payment request, any credential reset — verified on a known-good number or in person, never on the channel the request arrived through. Why: 66% of APP fraud starts online and the growth is entirely in authorised fraud, where the victim moves the money themselves. This control works regardless of how convincing the approach is. Write it as policy so a junior member of staff challenging a senior one is following a rule, not making an accusation. A callback protocol or shared phrase for finance and leadership. Agreed offline, used whenever a request involves money or access. Why: voice and video impersonation is the capability that has moved fastest, and it defeats “listen carefully” as a defence. Costs nothing, takes one meeting. Multi-factor authentication everywhere, with a named admin owner per account. Email, social, website, CMS, finance platform, supplier portals. Why: only 47% of businesses and 38% of charities have any 2FA, and takeovers of website, social media or email accounts hit 18% of FE and HE institutions. Most organisational account compromise traces to a former employee’s access or a shared password — an inventory problem, not a technology problem. Monthly self-search across the platforms your community uses. Your organisation name, your logo, your senior staff. Know the reporting route for each platform before you need it. Why: impersonation reached 79% of FE and HE institutions and 44% of secondary schools. Detection, not prevention, is the weak link — nobody can stop an account being created. Publish deliberately. Not less. But a directory of staff names, roles, direct emails and photographs is a convenience for legitimate contacts and a targeting package for everyone else. Why: personalisation is what moved the click rate from 5% to 17%, and personalisation is assembled from what you publish. Dormant account hygiene. Close what nobody uses, on a schedule. Why: one university in the survey described finding around 150,000 alumni accounts unused for years and closing them after they were used in attacks. The same principle scales down to the shared inbox nobody has opened since 2023. Train for reflexes, not definitions. Short, frequent, practical exercises using lures relevant to your own organisation: payroll, term dates, funder deadlines, invoice changes. Why: the 5% to 17% jump. Knowledge is not reaction under time pressure. One named owner reporting to the board or trustees at least termly. Not necessarily technical. Why: only 31% of businesses and 30% of charities have board-level responsibility assigned, and only 25% and 19% respectively have a formal incident response plan. Then Cyber Essentials, when you can. Certification follows the controls; it does not replace them. Why: 24% of businesses and 13% of charities already have the technical controls in all five areas, but only 5% and 3% are certified. Many organisations are closer than they think, and the assessment surfaces the gaps. 9. The first 90 days Nothing below requires new software, a consultant or a budget line. The only resource it needs is named time, which is why the first action is naming the person. Work through it in order: the early actions close the routes attackers are actually using, the later ones prove the earlier ones held. Days 1–30 · Close the open doors # Action Owner Time Done when 1 Name one person accountable for cyber security and record it Head / CEO 10 min Named in board or trustee minutes 2 Write and circulate the verification rule: no bank detail change, urgent payment or credential reset acted on without confirmation on a known-good number Named owner 1 hour One-page rule issued; finance has confirmed receipt 3 Brief the finance and leadership group in person and agree a callback phrase Named owner 30 min Phrase agreed and recorded offline 4 Build the account inventory: every email, social, website, CMS, finance and supplier portal account, and who holds admin on each Named owner 2–3 hrs Single list exists and is shared with the Head or CEO 5 Remove leavers and dormant accounts found in the inventory Named owner / IT provider 1 hour No active accounts for people who have left Days 31–60 · Harden what you own # Action Owner Time Done when 6 Multi-factor authentication enabled on every account on the inventory Named owner / IT provider 3–4 hrs Inventory shows MFA on, with exceptions listed and justified 7 First impersonation sweep: search your organisation name, logo and senior staff names across the platforms your community uses Comms or admin 45 min Findings logged; reporting route noted for each platform 8 One-page incident plan: who decides, who calls whom, who speaks to parents, supporters or customers, and when you notify the ICO Named owner 2 hours Plan issued and held somewhere reachable without network access 9 Confirm the backup actually restores, by restoring one file Named owner / IT provider 30 min Restore tested and dated Days 61–90 · Prove it works # Action Owner Time Done when 10 Run an internal phishing exercise using a lure relevant to you: payroll change, term dates, funder deadline, invoice update Named owner 2 hours Click rate recorded; no individual named in any report 11 Short, targeted follow-up for anyone who clicked — supportive, not disciplinary Named owner 1 hour Follow-up delivered within two weeks of the exercise 12 Check technical controls against the five Cyber Essentials areas and note the gaps Named owner / IT provider 2 hours Gap list written, with a cost estimate for anything chargeable 13 First report to the board or trustees: attempts seen, controls now in place, click rate, one thing you need Named owner 1 hour Item minuted, with a standing termly slot agreed Then, every term Re-run the impersonation sweep. Re-check the account inventory against your staff list. Repeat the phishing exercise with a different lure. Report all three to the board in one page. A note on the numbers Record your first click rate before you train anyone. It is the only baseline you will get, and it is the figure that makes the case at the next budget discussion. Expect it to be worse than you hope: one multi-academy trust in the government survey watched its own staff click rate move from 5% to 17% purely because the lure got better. What sits outside this plan Cyber Essentials certification carries a fee paid to the certification body, Microsoft 365 Business Premium carries a licence cost at charity rates, and managed backup carries a subscription. None of those are prerequisites for the ninety days above — but items 9 and 12 will tell you which of them you actually need, with evidence, before anyone asks for money. 10. How to put this to a board in three sentences The threat has industrialised, and the same infrastructure now serves both state influence operations and ordinary fraud, so capability reaches organisations like ours far faster than it used to. Government data shows 96% of breached secondary schools were phished, impersonation of institutions and their staff is rising sharply, and fewer than half of UK organisations have multi-factor authentication in place. The proposal is a named owner, a written verification rule for payments and credentials, MFA across every account, and a termly report — at no cost beyond time. If they ask what happens if you do nothing: the realistic outcome is not a dramatic breach. It is a redirected invoice, a fraudulent appeal in your name, or a supporter who paid money to an account that was never yours. Recoverable problems, and far cheaper to prevent than to explain. 11. Conclusion Small organisations were historically protected by irrelevance. Attacks were expensive to research and personalise, so attackers went where the money was. Personalisation is now nearly free. The research is automated, the content generated, the persona synthetic. The economics that kept a village primary school or a small charity below the threshold no longer hold — and the capability arriving at your door was developed by people whose objectives have nothing to do with you. The defences have changed far less. Verify out of band. Know what you own. Control who has access. Practise. Unglamorous, and what actually works against both halves of the same machine. Sources Department for Science, Innovation and Technology and Home Office, Cyber Security Breaches Survey 2025/2026 , published 30 April 2026. Fieldwork August to December 2025 by Ipsos; 2,112 businesses, 1,085 charities, 577 education institutions. DSIT and Home Office, Cyber Security Breaches Survey 2025/2026: Education Institutions Annex , published 30 April 2026. 273 primary schools, 222 secondary schools, 33 FE colleges, 49 HE institutions. Note: the education annex is not classified as official statistics and its samples are unweighted, with margins of error up to ±16.2 percentage points for FE colleges. UK Finance, Annual Fraud Report 2026 , June 2026. ENISA, Threat Landscape 2025 — 4,875 incidents, July 2024 to June 2025. European External Action Service, 4th Report on Foreign Information Manipulation and Interference Threats , March 2026. World Economic Forum, Global Risks Report 2026 . OpenAI, Disrupting Malicious Uses of AI , February 2026 update. Ofcom, Online Safety Act implementation — categorisation register and consultations published 10 July 2026. The Payments Association, The New Origin of APP Fraud , March 2026.