In this week’s guest column, behavioural scientist and author Dr Gleb Tsipursky, argues that AI transparency policies will only succeed when employees feel psychologically safe to disclose how they use AI at work. As new EU transparency obligations take effect, he explores why responsible AI governance must protect workers from blame while helping organisations build trust, accountability and responsible AI governance.
The European Union’s new artificial intelligence (AI) transparency obligations took effect on 2 August. They require clearer disclosure when people interact with AI and when certain synthetic or manipulated content reaches the public. These rules address an important question: when should people know that AI shaped an interaction or message?
Although the new obligations apply within the European Union, their influence is likely to extend much further. Many multinational organisations are already aligning AI governance across jurisdictions, making transparency and trust a global workplace issue rather than simply a European compliance requirement.
But responsible employers face a second question: what happens to the worker who discloses that AI shaped part of their work?
AI GOVERNANCE RISKS
Many organisations say they want transparency while maintaining a culture that punishes visible AI use. Employees hear that AI should improve productivity, but they also hear jokes about “letting the robot do your job.” Managers praise polished output while treating admitted AI assistance as evidence of laziness, weak judgment or diminished professional skill. Under those conditions, disclosure becomes personally risky.
The predictable response is concealment. Employees use approved and unapproved tools quietly, delete prompts, hide uncertainty and present outputs as wholly their own. This behaviour creates risks for data protection, quality and accountability. It also prevents organisations from learning which uses help workers and which merely transfer problems to colleagues.
As recently reported by Fair Play Talks, two-thirds (66%) of employees spend up to six hours each week correcting low-quality AI-generated work, while nearly half fix AI errors themselves rather than report or reject them, creating invisible labour that rarely appears in performance reviews. This growing phenomenon, sometimes described as “AI workslop”, demonstrates how poor AI governance can shift hidden costs onto employees rather than eliminate work.
TRANSPARENCY AND PSYCHOLOGICAL SAFETY
Ethical AI governance therefore needs a worker-protection dimension. Transparency depends on psychological safety. Employees must believe they can disclose AI use, question AI outputs and report problems without fear of embarrassment, punishment or damage to their professional reputation.
This is not simply a cultural aspiration. Fair Play Talks has previously highlighted research showing employees who feel psychologically safe are 2.1 times more motivated, 2.7 times happier at work and 3.3 times more likely to reach their full potential, while organisations that prioritise psychological safety also experience significantly stronger employee retention.
Transparency also depends on AI literacy. Employees need practical guidance on what constitutes appropriate AI use, when disclosure is expected and where human oversight remains essential. Without clear expectations, even well-intentioned employees may struggle to apply transparency consistently.
Disclosure should help organisations manage risk without turning employees into convenient targets when systems fail. Employers also have legitimate responsibilities to protect confidential information, comply with regulation and maintain confidence in AI-assisted decisions. The challenge is achieving those objectives without creating a culture of fear.
SEPARATING ACCOUNTABILITY FROM BLAME
Start by separating accountability from blame. Employees should remain responsible for checking their work and following clear rules. Yet responsibility does not require pretending that every AI failure reflects an individual moral defect. Poor outputs often reveal weak tool selection, vague policies, inadequate training, unrealistic workloads or missing review processes. Leaders should examine the system around the decision before assigning fault.
Ultimately, transparency is less about technology than organisational trust. When people trust that disclosure will be treated fairly, organisations gain better visibility into how AI is being used and where risks genuinely exist.
DEFINING MEANINGFUL AI DISCLOSURE
A generic instruction to “be transparent” leaves employees guessing. Organisations should distinguish routine assistance, such as outlining or editing, from higher-risk uses involving confidential information, hiring, performance evaluation, customer advice or public claims.
Workers need to know when they must identify the tool, preserve evidence, obtain review and inform the recipient. Disclosure should also be proportionate. Requiring a formal declaration for every spelling suggestion creates noise and encourages cynicism. Focusing on material influence gives employees a rule they can actually follow.
REWARDING HUMAN JUDGEMENT
Leaders must demonstrate that disclosure does not reduce professional status. A manager can openly explain that an approved tool produced a first draft while the manager verified the facts, applied context and rewrote the recommendation. This models a mature division of labour. The technology assists; the human remains accountable.
Front-line managers will ultimately determine whether transparency becomes routine or risky. Their response to the first disclosure of AI use is likely to shape how openly their teams engage with AI in the future.
Organisations should reward judgment. When performance systems value only visible effort or individual authorship, workers have an incentive to hide assistance. Evaluation should instead consider accuracy, quality, collaboration, risk awareness and the ability to improve a workflow. The employee who identifies a weak AI output before it reaches a client has created value, even when the first attempt failed.
SAFE CHANNELS FOR REPORTING RISK
Workers need a protected way to report AI problems. A confidential channel should allow them to flag hallucinations, biased recommendations, privacy concerns, unsafe automation and pressure to use a tool against policy.
Reports should feed regular reviews that improve training, approved-tool lists and human oversight. Unless someone deliberately conceals a serious violation, the first response should focus on learning and containment.
This approach also protects colleagues. Hidden AI use can generate “workslop“, where a quick output for one employee creates correction and verification work for others. Transparent teams can decide where AI saves time across the whole workflow rather than celebrating isolated productivity while shifting costs downstream.
Worker representatives and employee resource groups can also surface risks that formal governance committees overlook, especially when automation changes roles or redistributes invisible work.
TRUST: FOUNDATION OF AI TRANSPARENCY
The central ethical issue is not whether organisations disclose AI use in principle. It is whether their culture makes truthful disclosure possible. A policy that demands transparency while humiliating the person who complies will fail. It will drive AI use underground while preserving only the appearance of control.
Responsible leadership makes three commitments clear. Employees may use approved tools within defined boundaries. Humans remain accountable for consequential work. People can report problems without sacrificing dignity or professional standing.
The new European rules can help normalise disclosure in public-facing systems. Employers should extend that principle inward with equal care. The next phase of responsible AI governance will not be defined simply by what organisations disclose to customers or regulators. It will be defined by whether employees feel safe enough to be honest about how AI is really being used inside the workplace.
WHY THIS MATTERS
As AI becomes embedded across everyday work, transparency is becoming as much a leadership and workplace culture issue as a regulatory one. Fair Play Talks has previously reported that only 53% of European employees say they work in psychologically healthy organisations, highlighting the gap many employers still face in creating environments where people feel safe to speak up.
The issues raised in this article also connect with Fair Play Talks’ wider coverage of AI governance, workplace wellbeing and the future of work. As organisations accelerate AI adoption, success will depend not only on technology and compliance, but also on leadership, trust and cultures where employees feel confident using AI responsibly and speaking openly about its risks and limitations.
ABOUT THE AUTHOR
Dr Gleb Tsipursky is a behavioural scientist, CEO of Disaster Avoidance Experts and author of The Psychology of AI Adoption at Work: From Resistance to Results (Georgetown University Press, 2026). His research focuses on helping organisations adopt AI responsibly while building trust, resilience and effective decision-making in the workplace. Click here to find out more his book.
RELATED ARTICLES
Two-thirds of employees say they spend up to six hours a week fixing low-quality, AI-generated work, according to a new study.
A new study has revealed that only 53% of employees across Europe work in psychologically healthy environments.
One in three organisations are rehiring staff after AI-driven layoffs failed to deliver savings, highlighting hidden costs, lost productivity and workforce disruption.
A new study has confirmed the growing adoption of AI across industries, but also highlights a lack of responsible AI practices.
HR leaders believe business leaders are underestimating the people challenges of AI.
AI may be transforming workplace productivity, but a growing number of employees fear the technology is making them less intelligent in the process, according to a new survey.
Sign up for our newsletter
