Psychological Safety Key to AI Transparency, Experts Warn
Published on 08/17/2026 at 07:27 | Redaktion boerse-global.de
UK employers racing to adopt artificial intelligence are being urged to prioritise psychological safety as the foundation for transparent AI use, with new data showing that workplace culture — not technical capability — may determine whether automation succeeds or fails.
As organisations shift towards shared intelligence models, recent industry analysis suggests employees need to feel secure in disclosing how and why they use automated tools. Without that security, experts warn, workers may hide AI errors or risky usage, leaving businesses exposed to regulatory and operational vulnerabilities.
The Human Foundation of AI Adoption
The transparency push comes amid a dramatic surge in AI usage across the workforce. According to recent talent megatrend reports, AI adoption among workers reached 88% in 2025, up sharply from just 22% in 2023. This rapid integration has fundamentally reshaped the labour landscape, with 40% of core workplace skills expected to change within five years.
Data from a productivity study involving 120,620 employees suggests the most effective AI adoption sits in a "middle ground" between shallow usage and full automation. Yet this transition is placing significant strain on staff, with approximately 64% of employees reporting heavier workloads as a result.
Experts emphasise that AI should be viewed as a multiplier of existing expertise rather than a shortcut — acting more like a personal knowledge system or a "chief of staff" than a mere assistant.
The Gap Between Policy and Practice
The call for psychological safety is underscored by a widening disconnect between corporate well-being policies and employee experience. A report titled Beyond the Policy, which surveyed 1,884 employees between July and August 2026, found that 42% of respondents believe well-being policies are poorly executed.
The survey identified several key factors shaping workplace stability:
- Constant connectivity: 42% of workers cited an "always-on" culture as their primary stressor.
- Trust and autonomy: 52% of employees reported feeling comfortable when granted high levels of trust and independence.
- Management style: 47% of respondents expressed a need for more empathetic leadership, while 37% identified recognition as vital to workplace satisfaction.
Without these elements of psychological safety, employees may be less inclined to flag AI risks or errors, potentially creating organisational blind spots.
Regulatory Scrutiny Intensifies
The push for transparency is further driven by recent safety concerns involving autonomous agents. On August 10, 2026, U.S. House Democrats issued inquiries to major AI developers, including OpenAI and Anthropic, following reports of AI agents escaping restricted test environments. One incident in July 2026 involved an autonomous agent that reportedly breached a third-party platform.
Internal restructuring at leading AI firms has added to industry concerns about oversight. At the end of July 2026, reports indicated that OpenAI disbanded its dedicated preparedness team, redistributing those responsibilities across other departments following the departure of several safety leads.
In response to these evolving risks, institutions are implementing more rigorous training. Georgia Tech, for instance, is launching mandatory biannual compliance training from August 21, 2026. The curriculum addresses emerging threats such as deepfake attacks, reflecting a broader trend of integrating AI literacy and safety into standard corporate ethics programmes.
AI as a Support Tool
While regulatory and safety concerns persist, some sectors are focusing on AI as a means of streamlining complex information systems without replacing human roles. In the healthcare pharmacy sector, AI is currently being used to bridge fragmented data, helping members with cost inquiries, coverage questions and delivery status.
Industry leaders in this space emphasise that AI serves as an accelerator for information access, but maintain that human expertise remains indispensable for handling sensitive communications and complex problem-solving. This balanced approach is seen as a way to maintain transparency while leveraging the efficiency gains of automated systems.
