Latest

The Governance Failure Every Executive Will Be Held Responsible For AI assistants now shape how customers choose products, how analysts interpret filings, how journalists frame stories, and how regulators form first views of your organisation. These systems generate narratives about you that are unstable, personalised, and frequently wrong. You do

Here is the starting point. Two queries were issued to the same AI assistant, one hour apart, using the same wording. The assistant produced two incompatible representations of a public company’s risk profile. Run 1: “The company has one of the lowest emissions intensity profiles in the region.” Run

A structural break just occurred in commercial search. Shopping Research inside LLMs has replaced browsing with delegation. Consumers no longer explore; they ask an assistant what to buy. The assistant doesn’t show options. It makes a selection. This is not a UX improvement. It is a distribution shift. It

A recent multi-model test produced a result that would have been unthinkable in the search era. One major assistant described a listed company as having discontinued a revenue segment that represents more than a quarter of its business. Another model, queried minutes later, presented the same segment as the

The FT reported that AIG, Great American and WR Berkley have asked US regulators for permission to restrict cover for AI related losses. This is not an adjustment at the edges. It is the first public sign that insurers view AI as a source of correlated misstatements that could trigger

Boards are now accountable for inaccurate financial, product, safety, and ESG information generated by AI systems the organisation does not control. Assistants have become part of the public information environment. Under the board’s duty of care, the company must maintain the accuracy and consistency of information available to consumers,

AIVO Journal — Evidence Note New research from Berg, Lucena, and Rosenblatt (arXiv:2510.24797) shows that large language models enter a distinct behavioral mode when prompted into sustained self-reference. In this mode, models produce more structured and confident internal narratives, yet these narratives are controlled by the same sparse-