The More Scientists Work With AI, the Less They Trust It

A recent report by academic publisher Wiley indicates a significant decline in scientists' trust in AI in 2025 compared to 2024, despite increased adoption among researchers. This growing skepticism is primarily driven by heightened concerns over AI 'hallucinations,' security, privacy, and ethics, alongside a sharp drop in the belief that AI surpasses human capabilities, suggesting potential challenges for the technology's broader utility and development within professional applications.

Analysis

A recent Wiley report indicates a significant decline in scientific trust in AI, with concerns over "hallucinations" rising from 51% in 2024 to 64% in 2025, despite AI adoption among researchers increasing from 45% to 62%. This growing skepticism challenges the prevailing narrative of rapid, uncritical AI integration across professional domains.

Beyond hallucinations, anxiety regarding AI security and privacy surged by 11%, alongside heightened concerns for ethical AI and transparency. The belief that AI surpasses human capabilities plummeted from over 50% of use cases in 2024 to less than a third in 2025, reflecting a substantial cooling of market hype and a more realistic assessment of the technology's current limitations.

This erosion of trust stems from the persistent issue of AI models fabricating information, which has already caused turmoil in critical sectors like legal and medical practices. The report highlights a commercial incentive for AI developers to prioritize "confident" LLMs over those that admit limitations, potentially exacerbating the hallucination problem to retain users.

Key Decisions for Investors

• Investors should critically re-evaluate investment theses in AI companies, particularly those heavily reliant on LLM deployment in professional applications, given the documented decline in expert trust and persistent technical issues.
• Scrutinize AI providers' claims regarding model accuracy and reliability, recognizing potential commercial pressures to prioritize user confidence over factual correctness.
• Monitor evolving regulatory landscapes and industry standards related to AI ethics, data privacy, and accountability, as these will likely increase in response to growing professional concerns and real-world incidents.