When we’re talking about high-stakes online exams, the stakes aren’t just high for the students.
In higher education exams that determine entry into careers such as medicine, law, and engineering, the stakes are also high for the public, and for the institutions that certify graduates’ competence.
Academic leaders are increasingly concerned by the strategic threat posed by AI cheating in online exams. The institutional risk they introduce cuts across financial sustainability, accreditation standards, and wider reputation.
In this five-minute briefing, we’ll discuss:
- The scale of AI cheating and associated institutional risk
- The direct and indirect costs of AI cheating to institutions
- Strategies to protect academic integrity in high-stakes exams
However, beyond these strategies, we must confront a new technological reality: as long as an exam platform requires an internet connection to function, it leaves a digital backdoor open for AI interference. The strategies that protected exams yesterday are simply architecturally insufficient against today’s OS-level AI tools.
How significant is the risk of AI cheating? Evidence from the UK
In the UK in June 2025, The Guardian used Freedom of Information responses from 131 UK universities to quantify the scale of the problem. They discovered nearly 7,000 cases of AI-related cheating during the 23/24 academic year. This equalled 5.1 cases per 1,000 students, up from 1.6 the previous year.
The report goes on to say: ‘Figures up to May suggest that number will increase again this year to about 7.5 proven cases per 1,000 students – but recorded cases represent only the tip of the iceberg, according to experts.’
Meanwhile, research from the UK’s Higher Education Policy Institute with Scottish universities found that instances of academic misconduct investigations have increased, in line with the launch of generative AI, though the rate varies from 75% all the way up to 411% increase.
Between the financial, reputational, and accreditation risks, AI cheating comes at a high cost for higher education, and the evidence suggests this cost has the potential to grow if institutions fail to curb AI-enabled misconduct in assessments.
What makes AI cheating an institutional risk?
AI cheating is an institutional risk – rather than just an academic integrity or teaching issue – because it has the potential to undermine reputation and revenue drivers for organizations.
AI-enabled misconduct devalues degrees, and those devalued degrees can impact accreditation, applications, professional partnerships, and funding opportunities.
There are also direct financial risks. Without a strategic leadership response, institutions face escalating costs associated with investigating misconduct and increased legal exposure relating to student challenges.
Let’s explore the five institutional risks of AI cheating in more depth.
1. The financial risk of AI cheating
Analyzing over 2,000 cases since 2019, Higher Education Policy Institute research estimates the cost of investigating academic misconduct to be £12.4 million in the UK and $196 million in the US.
They determined that each case takes 56 minutes of academic time and 106 minutes of administrator time to resolve. Based on the Scottish pay scale for each role and an assumption of investigating 1,000 breaches per year, this represents a total annual cost of £95,181 per institution, equivalent to over 950 postgraduate scholarships that could have been funded.
This represents a significant financial burden for institutions, as well as putting additional pressure on faculty teams to conduct investigations. Given that the cost of turnover is estimated to be 33% of departing staff salary, burnout and AI fatigue are additional indirect financial factors to consider.
Furthermore, IT and academic teams face an endless, costly cycle of trying to block new AI apps and browser extensions as they appear. This reactive 'whack-a-mole' approach drains resources and budget without ever truly securing the exam environment against the next update.
2. The legal risk of AI cheating
Other potential costs arising from AI cheating could include: external legal representation for appeals, expanding compliance teams, and legal settlements or damages.
Institutions that fail to address – or successfully prove – AI misconduct may face appeals, disputes, or even litigation from students, requiring external counsel, settlements, or extended administrative resources.
Norway has recently put a cap on the legal fees students can reclaim from their universities, to deter appeals that had become ‘time-consuming and expensive beyond reason’ after a spike in appeal cases following the introduction of remote exams during COVID. While cases have since dropped in the country, this move speaks to the drain that legal action represents to institutional resources.
3. The accreditation risk of AI cheating
Accreditation bodies rely on institutions to uphold rigorous academic standards, including integrity in assessments. AI cheating in exams directly threatens those accreditation standards by undermining the reliability of exam results, risking under-prepared professionals entering the workplace bearing their seal of approval.
If institutions can’t protect academic integrity, accrediting organizations may choose to withdraw their support from your programs, which can lead to declining enrolments.
Accreditation standards and audits include checks on exam security and integrity processes. Failure to demonstrate proactive measures against the threat of AI-enabled misconduct may result in warnings or even suspension of accreditation.
4. The reputational risk of AI cheating
Higher education reputation directly translates into financial sustainability. When your reputation precedes you, opportunity follows: student and graduate applications, employer and research partnerships, accreditation, influence, and more.
If an institution builds a reputation for failing to curb academic misconduct, every student and graduate faces doubt about the validity of their qualification. This can impact future applications.
Plus, like accreditation bodies, no institution wants under-qualified professionals announcing them as their alma mater. This erodes public trust, student applications, employer relationships, and more – all of which contribute to overall institutional viability.
5. The academic risk of AI cheating
Beyond the high-level institutional damage, this issue also has a direct impact in the classroom. The primary concern is that AI cheating masks gaps in student knowledge and makes pass rate data unreliable.
When educators can’t accurately identify where students are struggling, this undermines their ability to provide effective and timely remediation. It also conceals potential issues in program design. Both of these can compromise the quality of student and graduate outcomes.
Without a secure baseline, institutions cannot distinguish between a student who has mastered the material and one who has mastered the art of prompting, and this can impact student remediation.
How academic leaders can protect against AI cheating in digital exams
1. Consider direct and indirect institutional costs
Beyond calculating the direct financial cost of misconduct, leaders should evaluate the broader impact on accreditation, graduate outcomes, and public trust. Conduct scenario analysis to understand how AI cheating could affect applications, funding, and partnerships over time.
2. Redesign assessments strategically
Research finds that AI-detection-only policies are inadequate and HE needs ‘fundamental pedagogical realignment’ to ‘validity-driven assessment practices’ to mitigate AI risk (Leaton Gray, et al, 2025).
Follow best practice from global education leaders by redesigning assessments to deter AI misconduct, especially as browser and OS-level artificial intelligence has rendered once-secure digital exams vulnerable to AI.
To do this, it's important to consider the common denominator when it comes to the source of AI. And that's the internet.
By using an offline delivery platform, such as ExamSoft by Turnitin, institutions can create a secure ‘air gap’ that locks down the entire device. This closes the internet backdoor that allows students to access OS-level AI (e.g. Microsoft Copilot) and give access to AI agents that can control their device.
3. Teach AI literacy and academic integrity
Promote responsible AI use to help students develop the skills they need to incorporate AI into their learning workflows constructively and ethically.
50% of students are confident doing this, but this means the other half aren’t. Research finds that ‘the majority of institutions do not have formal guidelines for AI use, leading to confusion among students and instructors.’ They should therefore prioritize policies for ‘equitable, inclusive, and ethical use of AI’ (Song NaYoung, 2024).
4. Combine internal oversight and external expertise
AI is evolving rapidly, and keeping pace can be challenging for any institution. Designate a leader to monitor trends in AI-enabled misconduct and collaborate with technology partners who specialize in staying ahead of emerging risks. By combining internal oversight with external expertise, institutions can proactively address AI challenges and safeguard their reputation.
The true cost of AI cheating: Worrying but preventable
AI-enabled misconduct in high-stakes online exams carries costs that go far beyond the classroom.
Direct financial costs include staff time investigating cases, compliance activities, and potential legal expenses from student challenges. These can amount to tens of thousands of pounds per institution each year.
Indirect costs are equally significant. Devalued degrees that threaten accreditation, damage higher education reputation, and undermine public trust – all of which can impact applications, partnerships, and funding opportunities.
Education leaders are addressing these costs by redesigning assessments, swapping from online to offline digital exams, building AI literacy across the institution, and making AI integrity part of the curriculum. This not only protects academic integrity but also reduces the financial, reputational, and legal exposure that the AI era brings to higher education.
Ultimately, the most effective way to protect the value of your credentials is to change the architecture of your security. By disconnecting digital exams from the internet, you don't just detect cheating, you prevent the opportunity for it to happen.
Close the AI-sized gap in your digital exam security.
Discover how ExamSoft by Turnitin creates a secure, offline environment that neutralizes OS-level AI threats and protects the integrity of your high-stakes exams.