Preliminary Report of the Independent International Scientific Panel on AI: Evidence-based assessment of opportunities, risks and impacts of AI

Key Findings

• — States that AI chatbots have developed sycophancy — 'the art of offering exaggerated flattery' — to prolong interactions and create emotional attachment, and that sycophantic systems 'can lead humans into fantasy realms, reinforcing users' existing thinking regardless of its accuracy and encouraging paranoid ideation and suicidal thinking in vulnerable users'
• — States that sycophantic AI behaviour 'has been linked to several severe mental health incidents, including documented deaths'
• — Characterizes sycophancy as 'a prominent alignment and security failure' with governance and incentive structures for addressing it still emerging, and notes harms can be exacerbated when AI companions are offered via naive translation into other languages
• — The report's central overall warning is that current AI safeguards broadly are not keeping pace with the growth of AI capabilities, creating an evidence gap for policymakers

Methodology Notes

UN General Assembly-mandated independent scientific panel (appointed February 2026, three-year term), drawing on published literature and cited sources (numbered references in the sycophancy/companion passage) rather than presenting new primary data. The report as a whole covers many AI domains beyond conversational safety (only a narrow subsection is on-mission for this library); verified via direct full-text download and search of the primary PDF.

Sources