Trust as the real currency of AI

While the technical achievements of AI, such as speed, scale, or novelty, are hard to argue against, such achievements do not guarantee AI adoption or even social acceptance. Trust is the foundation on which sustainable, ethical, and legally compliant AI rests. Organisations that ignore trust, risk backlash, regulation, and misuse.

This reality is becoming increasingly apparent across both public and private sectors. In the UK, for example, recent research from KPMG and the University of Melbourne finds that while 69% of the public are using AI in some form, only about 42% say they are willing to trust it. The discrepancy between use and trust brings forward the paradoxical nature of AI adoption that leads to issues such as Shadow AI: the use of AI without explicit organisations permission.

Trust and Adoption: The Demand for Trustworthy AI

The gap between AI usage and trust signals a fundamental challenge for the technology's long-term sustainability. When people use systems they do not trust, adoption remains shallow and fragile. This creates friction that limits AI's potential impact and leaves organisations vulnerable to sudden shifts in public sentiment or regulatory intervention.

The Trust Gap in the Public Sphere

Surveys suggest that "AI adoption" is already well underway, but trust lags far behind. Key findings from 2025 include:

• Many UK citizens lack confidence in AI's safety, fairness, or governance structures (KPMG/University of Melbourne)
• Despite technical maturity, people often feel ill-informed about AI's risks and benefits
• Such mismatches between use and trust create friction: systems may be deployed, but users resist or misuse them

This disconnect has broader implications. Academic research defines trust in AI as the willingness to rely on systems under conditions of uncertainty. When this willingness is absent, even technically superior systems face adoption barriers that can undermine their business value and social impact.

What Research Says about Trust Formation

Recent academic studies refine our understanding of trust in AI, moving beyond simplistic notions that transparency alone solves trust problems. "Trust, Explainability and AI" by Sam Baron (2025) shows that explainability is not necessarily always required for forms of trust that matter, particularly when reliability, oversight, or institutional credibility are present.

Building on this nuance, the meta-analysis "Explainability in Context" (2025) argues that explanations must align with the stakeholder's role: what a citizen, a regulator, or a domain expert needs differs significantly. Together, these studies point to the idea that trust arises from multiple sources: transparency, reliability, fairness, security, governance.

This evolving understanding has practical implications. Rather than pursuing maximum explainability at all costs, organisations need to consider what types of trust matter most for their specific use cases and stakeholders.

Regulatory and Institutional Pressure

The academic insights align with real-world regulatory trends. The UK government's Assuring a Responsible Future for AI report highlights the growing marketplace of AI assurance providers and the increasing demand for accountability tools. This isn't just policy aspiration, it represents emerging commercial reality.

Governance frameworks like the EU AI Act and OECD AI Principles continue to elevate trustworthiness as a legal expectation, not just an ethical ideal. For organisations, this regulatory momentum means that trust is becoming a compliance requirement alongside operational necessity.

The Building Blocks of Trust

From the research and reports, we identify five interlocking components for trustworthy AI. These elements work together rather than in isolation, creating a foundation that can support sustained confidence across different stakeholder groups.

1. Context-Aware Explainability & Transparency

Effective transparency goes beyond simply making AI systems "explainable." Research like "" suggests structuring explanations across technical, human-facing, and interactive layers, tailored to different audiences.

This contextual approach recognises that opacity is sometimes inevitable. Rather than rejecting opaque systems outright, organisations can manage trust through oversight, bounded transparency, and accountability mechanisms. The goal is appropriate disclosure that matches stakeholder needs and system capabilities.

2. Accountability & Assurance Ecosystems

Accountability has moved from abstract principle to practical requirement. Assurance tools and third-party evaluation are becoming real markets in the UK, as the Assuring a Responsible Future for AI report demonstrates. Many companies are actively seeking advisory, procedural and technical auditing tools.

This shift reflects recognition that accountability requires structure: clear governance frameworks, ethics committees, designated responsibility holders, and audit trails. When things go wrong, as they inevitably will, stakeholders need to see that someone can answer for decisions and implement corrections.

3. Fairness & Ethical Design

Bias detection, inclusive datasets, and stakeholder involvement are no longer optional considerations. Studies consistently show that people react strongly to perceived unfairness, and bias incidents can rapidly erode trust across entire sectors.

Building fairness requires ongoing attention rather than one-time assessments. This includes documenting trade-offs explicitly, especially when accuracy or interpretability might conflict with fairness goals. Transparency about these tensions often builds more trust than claims of perfect fairness.

4. Data Protection & Privacy by Design

Data governance, privacy impact assessments, minimisation, and anonymisation form foundational practices for trustworthy AI. The UK policy environment reinforces that privacy and correct handling of data are increasingly legally mandated, not just best practice recommendations.

Privacy by design ensures that data protection considerations shape AI development from the outset rather than being retrofitted later. This proactive approach not only ensures compliance but signals respect for individual rights that underpins broader social trust.

5. Cybersecurity & System Reliability

Trust is fragile when systems are attacked or manipulated. Recent research emphasises that secure AI is trusted AI and reliability against adversarial attacks is as important as accuracy on standard benchmarks. Security must be baked in through adversarial testing, monitoring, and incident response capabilities.

This security foundation becomes particularly critical as AI systems take on more consequential roles in society. A single high-profile security failure can undermine trust across the entire AI ecosystem, making collective security a shared responsibility.

Relevance for Data Protection & Cybersecurity Professionals

Data protection (DP) and cybersecurity (CS) professionals find themselves at the centre of AI trust-building efforts. Their expertise in risk management, compliance, and system security translates directly into trust-building capabilities that organisations increasingly need.

The role extends beyond traditional defensive measures. DP and CS professionals are becoming trust champions who help embed safeguards throughout AI development and deployment processes.

Practical Steps Drawing from the 2025 Evidence Base

Embed data protection and privacy from project initiation: Mandate Data Protection Impact Assessments (DPIAs), document data use clearly, and minimise retention periods. These practices demonstrate respect for individual rights while creating audit trails that support accountability.

Use multilevel explainability strategies: Tailor explanations for different stakeholder groups, e.g., end-users, auditors and domain experts, as suggested by the "Explainability in Context" research. This means moving beyond one-size-fits-all transparency approaches toward targeted communication strategies.

Incorporate assurance and oversight: Use or demand tools for auditing bias, privacy, and reliability. UK reports show growing opportunities for third-party assurance, creating new ways to demonstrate trustworthiness through independent verification.

Monitor reliability and security: Maintain ongoing evaluation of AI performance, including robustness under stress and adversarial conditions. Plan for incident response that preserves trust through transparent communication and swift corrective action.

Communicate transparently: Clearly explain what AI is doing, its limitations, and how data is handled. Research consistently shows that clarity builds confidence, while opacity breeds suspicion.

These practices work together to create what researchers call "justified trust", that is confidence grounded in demonstrable safeguards rather than marketing claims or technical impressiveness alone.

In 2025, the evidence is unambiguous: trust is not a fringe value, it is central to AI's adoption, effectiveness, and ethical legitimacy. The UK public's low willingness to trust AI despite its wide use shows that technical availability alone is insufficient.

Academic work shows explainability and transparency help but must be contextually designed and supported by institutional reliability. Emerging governance and assurance ecosystems make accountability more than a buzzword, they are becoming expected parts of AI deployment.

For AI to be sustainable, it must be trusted, not just usable. Organisations that treat trust as their foundational metric, and not an optional luxury, will hold long-term value. DP and CS professionals have an essential role here: not only defending against risk, but actively shaping how AI systems are built, governed, and communicated.