The European Union’s General Data Protection Regulation (GDPR) and the AI Act represent landmark efforts to regulate algorithmic decision-making. Articles 15 and 22 of the GDPR grant individuals the right to meaningful information about the logic involved in automated decision-making (Regulation (EU), 2016; Malgieri and Comandé, 2017).

Article 15, the right of access by the data subject, is particularly relevant.³ It grants individuals the right to obtain confirmation as to whether personal data concerning them is being processed, and if so, to access that data and specific information about the processing. Article 15(1)(h) states that this information includes “meaningful information about the logic involved, as well as the significance and the envisaged consequences of such processing for the data subject.” This provision seems to directly address the need for individuals to understand how automated decisions affecting them are made.

Furthermore, Article 22 grants data subjects the right “not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning him or her or similarly significantly affects him or her.” While this article provides a right against solely automated decisions in high-stakes scenarios, it implicitly underscores the need for transparency and human involvement where such decisions are permitted or used to inform human decisions.

However, these promises of transparency are significantly undermined by countervailing protections for corporate interests. Recital 63 of the GDPR explicitly states that the right of access “should not adversely affect the rights or freedoms of others, including trade secrets or intellectual property”. ⁴

In practice, this exception is often used to block any meaningful access (Wachter et al., 2017). People trying to obtain greater algorithmic transparency under the GDPR have frequently encountered this trade secret defense from companies, highlighting the practical difficulties individuals face in exercising their rights when confronted with powerful corporate interests (Veale et al., 2021).

When a driver asks their telematics insurer why their premium has increased, they are likely to receive vague categories of risk factors (driving style, route choices) rather than the specific weights and data points that constitute the meaningful information they need (Bucher, 2017). The algorithms remain a black box. This creates a regulatory paradox where the right to transparency exists in theory but is often unenforceable in the face of corporate claims of confidentiality (Veale et al., 2021).

The European Union’s Artificial Intelligence Act (AI Act) represents a first comprehensive legal framework.

A framework specifically designed to regulate AI systems.⁵ Adopting a risk-based approach, the AI Act imposes varying levels of obligations on AI systems depending on their potential to cause harm. AI systems used in critical areas such as healthcare, education, employment, and law enforcement are classified as high-risk and are subject to stricter requirements.

The AI Act includes specific transparency obligations for high-risk AI systems under Article 13 and broader transparency obligations for certain AI systems under Article 52. For high-risk systems, providers must design and develop them to enable human oversight and provide accompanying information that is “clear and adequate.” This information is intended to help users understand the system’s capabilities and limitations.

Article 52 imposes specific transparency obligations for AI systems intended to interact with individuals or generate content, requiring users to be informed that they are interacting with or exposed to AI.

While the AI Act reinforces the importance of transparency and accountability for high-risk AI, it does not fully resolve the problems we face with implementing the GDPR in a way that protects trade secrets (Oxford Law Blogs, 2025). While it builds upon GDPR principles, the AI Act’s focus is more on the safety and fundamental rights risks posed by the AI system itself, rather than the data protection rights of individuals concerning automated decisions.

The Act aims to balance innovation with safety and rights, but the specific details of how transparency will be enforced and how it will be weighed against claims of confidentiality remain subject to ongoing debate and the development of implementing standards and guidelines.

The burden often remains on affected individuals or civil society organizations to identify problematic AI systems and advocate for greater transparency and accountability.

Furthermore, the regulatory landscape for AI transparency is fragmented. Even with the GDPR and AI Act, individuals seeking to understand how algorithmic decisions are made about them must navigate complex interactions between data protection law, sector-specific regulations (e.g., in healthcare or finance), and intellectual property law (Mittelstadt, 2019).

The lack of a single, clear, and universally enforceable right to truly understand algorithmic logic significantly hinders the ability of individuals to exercise agency and contest decisions that impact their lives, contributing to the opacity that underpins the reproduction of inequality.⁶

To understand the persistent resistance to transparency, we must look deeper than legal texts. Causal Layered Analysis (CLA), a futures research method, helps deconstruct the narratives that sustain the status quo (Inayatullah, 1998). CLA is a poststructuralist futures research method that moves beyond conventional, surface-level analyses of issues to uncover the deeper causes, worldviews, and metaphors that shape our understanding and limit possibilities for change (Inayatullah, 1998). By applying CLA to the debate surrounding algorithmic transparency and the fear of providing access and transparency, we can peel back the layers and reveal the often-hidden power dynamics and ideological commitments that maintain the status quo of algorithmic opacity (Figure 1).

CLA operates on four distinct, yet interconnected layers:

At the surface level, the litany of the algorithmic transparency debate is dominated by immediate concerns about the potential negative consequences of granting individuals access to the inner workings of AI systems. The narrative frequently presented in media, policy discussions, and by corporate actors focuses on the risk that users will game the system if they understand its logic (Van den Boom, 2020). Examples cited include individuals manipulating credit scoring algorithms to improve their rating without genuine financial responsibility, drivers altering behavior only when telematics systems are active, or students finding loopholes in educational assessment AI.

This layer emphasizes threats to algorithmic integrity, system security, and the potential for widespread manipulation or exploitation. The proposed solutions at this level often involve technical safeguards to prevent gaming, legal penalties for misuse, or simply maintaining secrecy to make gaming impossible. This litany, while containing elements of valid concern about system security, tends to frame the issue as a problem of malicious individual behavior that must be controlled.

Moving beneath the surface litany, the systemic layer reveals the underlying social, economic, and legal structures that give rise to and sustain the fear-of-gaming narrative. A primary systemic cause is the economic incentive for companies to protect their AI algorithms as valuable trade secrets and intellectual property. In a highly competitive market, the specific design, training data, and operational parameters of a performant algorithm can represent a significant competitive advantage. Transparency is perceived as a direct threat to this advantage, potentially allowing competitors to replicate successful models without incurring the same research and development costs (Van den Boom, 2020). This economic structure creates a powerful, vested interest in maintaining algorithmic opacity, and the legal frameworks surrounding intellectual property often provide robust mechanisms for doing so, creating a direct conflict with data protection and transparency rights.

Furthermore, the concentration of power within large technology companies and institutions that develop and deploy AI systems is a key systemic factor. This concentration of power allows these actors to shape the discourse around AI, influencing policy and public perception.

The fear of gaming can be strategically amplified by those in power to justify maintaining control and limiting external scrutiny. The complex technical nature of advanced AI also acts as a systemic barrier, creating an information asymmetry between those who build and deploy AI and those who are subjected to its decisions, reinforcing existing power inequality. The fragmented and often weakly enforced regulatory landscape for AI accountability also contributes to this layer by failing to create sufficient systemic pressure for meaningful transparency.

The systemic causes are, in turn, supported by deeper worldviews and paradigms. A dominant worldview underpinning the fear-of-gaming narrative is a strong belief in control and order imposed from the top down. This worldview views systems as entities to be managed and protected by experts and authorities, with users as passive recipients or potential disruptors who need to be controlled or contained. From this perspective, granting individuals the power that comes with understanding a system’s inner workings is inherently risky and undesirable, as it might lead to unpredictable outcomes outside of intended control.

Another powerful worldview at play is the prioritization of market dynamics and corporate interests over individual rights and democratic accountability. This perspective holds that the pursuit of economic efficiency, innovation (often narrowly defined by market advantage), and corporate profitability are paramount. Within this worldview, the protection of trade secrets is seen as essential for market functioning and innovation, and concerns about individual transparency rights are secondary or viewed as obstacles to progress. This worldview often assumes that market competition will eventually lead to optimal outcomes, including trustworthy AI, without the need for extensive external regulation or mandatory transparency that might impede corporate strategies. This perspective often downplays or fails to adequately account for how market forces can exacerbate, rather than mitigate, inequality and social harms when left unchecked (Whittaker, 2021).

At the deepest layer, the worldviews are sustained by powerful, often unconscious, myths and metaphors. The fear of gaming taps into deep-seated cultural myths about human nature, often portraying individuals as fundamentally self-interested actors prone to cheating and exploiting systems for personal gain. This myth justifies the need for external control and surveillance, reinforcing the idea that individuals cannot be trusted with power or knowledge about the systems that govern them.⁷

At the same time, there are powerful myths surrounding AI itself. AI is often portrayed metaphorically as objective, a neutral judge capable of making perfectly rational decisions free from human bias (Gillespie, 2014). We also see the narrative that AI systems are too hard to understand even for expert, so we just have to accept instead of scrutinize these systems.⁸ These myths contribute to a sense of technological determinism or inevitability best left in the hands of a select few experts (Morozov, 2013; Broussard, 2019; De Liban, 2024). The combination of the myth of the untrustworthy individual and the myth of the inscrutable or infallible algorithm enables the narrative that denying transparency is necessary and appropriate (Schellmann, 2024; Kroll et al., 2017). The algorithm becomes a benevolent, necessary authority that must be protected from the potentially malicious actions of individuals seeking to exploit its secrets (Kak and West, 2023).

Using Causal Layered Analysis reveals that refusing transparency and access rights is not a simple response to security risks. It is a narrative deeply embedded in structures of power, competition, and worldviews that prioritise corporate control over individual agency (Van den Boom, 2022). Challenging this requires more than demanding transparency; it requires a fundamental reframing of what it means to interact with an algorithmic system (Rudin, 2019).

In everyday use, gaming the system implies deception or manipulation. We acknowledge this cultural framing. However, in the context of opaque, high-stakes AI systems, often shielded from scrutiny by trade secrets or proprietary protections, gaming becomes a rational and necessary response to structural power imbalances (Kak and West, 2023; Edwards and Veale, 2018). When official channels for redress or explanation fail, users are left with few options but to experiment, test, or subvert the system to understand or challenge it.

We argue that gaming in this context takes on multiple democratic functions:

This reframing aligns with the growing literature on contestable AI, which seeks to provide procedural mechanisms for users to intervene in automated decision-making (Wachter et al., 2020). Our concept of the right to game AI systems is a speculative legal proposition that extends beyond explanation rights, advocating for user-driven practices of resistance, redress, and reappropriation.

This re-appropriation finds further grounding in the legal recognition of ethical hacking, red teaming, and adversarial testing as legitimate modes of system critique (European Union., 2024; Veale, 2020). However, these practices are usually reserved for technical experts under institutional oversight. Our speculative intervention envisions a future in which such practices are democratized. Here, the individuals most impacted by algorithmic systems (welfare recipients, insured drivers) are empowered to act as civic auditors, drawing on their lived experience to test, probe, and challenge the logic and effects of these systems (Van den Boom, 2023).

In this framing, gaming shifts from a self-interested or deceptive practice into a distributed method of algorithmic accountability. It functions as a form of grassroots red teaming, allowing users to stress-test decision systems and demand more robust, just, and transparent design. Unpacking these socio-technical infrastructures requires interventions from multiple, distributed vantage points (Crawford and Joler, 2018).

While the notion of contestability has gained traction in AI governance discourse, it often remains tethered to formal legal procedures or institutional processes (Wachter et al., 2020). We retain the term gaming because of its speculative, insurgent potential. Contesting suggests recourse within an existing system; gaming implies tactics deployed precisely when those systems are inaccessible, incomplete, or untrustworthy.⁹

The Right to Game AI Systems is thus presented not as a formal legal right in the traditional sense, but as a speculative legal artifact, a tool to provoke discussion about what rights might be needed when we are governed by inscrutable, non-negotiable infrastructures. It reflects the critical legal insight that law itself is often structured to deny access or recognition to certain subjects, and that resistance must often come from outside its formal channels (Delacroix and Wagner, 2021).