The Evidence: A Record of Observed Behaviour in External AI Systems (www.aivojournal.org)

A new evidentiary record has been established to document the observable behaviors of external AI systems, which significantly impact enterprise decision-making by shaping purchasing choices, risk assessments, and reputational trust. Conducted over six months in 2025, the study analyzed over 3,700 pages of AI-generated outputs from various enterprises in regulated and non-regulated sectors. This comprehensive corpus highlights recurring issues, such as AI systems substituting entities without disclosure, omitting relevant businesses in recommendations, and generating inconsistent outputs across different models and time frames. The significance of this record lies in its documentation of structural gaps in governance and oversight concerning AI-mediated representations. As end-users retain evidence of AI interactions, enterprises often lack access to any record of their portrayal, presenting challenges during disputes or regulatory scrutiny. The study underscores the urgent need for robust correction mechanisms and governance frameworks, particularly in industries where misrepresentation can lead to serious repercussions. This foundational corpus serves as a reference point for ongoing discussions about the accountability and regulation of AI systems, highlighting the necessity for a systematic response to these behaviors.