White Papers
Original thinking on the behavioral side of AI.
Behavieural's white papers set out the argument: where AI governance, oversight and adoption assume behavior that doesn't occur, and what to do about it. Each draws on research in behavioral science, psychology and human factors.
The white papers argue. The frameworks measure. A paper makes the case that something fails for behavioral reasons. A framework is the instrument for finding out whether it's happening in a specific organization.
-
Behavioral Governance
Engineering governance structures that make intended behavior the path of least resistance.
The argument. Organizations have invested heavily in model validation, cybersecurity, data governance and compliance, yet AI deployments still fail at the point of human use. The cause is often not the algorithm. Governance structures assume people will behave as policies intend, an assumption rarely tested.
The paper introduces behavioral governance as the layer above the five traditional pillars (technical, compliance, ethical, operational and lifecycle). Those pillars define how AI should be used. Behavioral governance tests whether the structures produce the intended behavior under real operational pressure. It treats behavior as a predictable output of institutional design, including decision rights, escalation pathways, accountability boundaries, workflow constraints, incentives, friction and social norms.
What's inside. A critique of the five governance pillars, the theoretical foundations, how the AI Trust Axis operationalizes the discipline, case scenarios and a five-level maturity model.
Who it's for. Governance, risk and compliance leaders, executives and boards responsible for AI oversight.
-
Governing Agentic AI
A behavioral science framework for governing the human–agentic AI relationship.
The argument. Agentic systems execute extended sequences of actions with only intermittent human oversight. The governance challenge shifts from evaluating isolated AI decisions to ensuring that people can supervise, understand and intervene in autonomous processes operating over time.
The paper identifies five behavioral mechanisms that shape that relationship: automation bias, responsibility diffusion, anthropomorphic trust, the autonomy paradox and failure recovery. It argues they are central determinants of whether agentic deployments stay safe, accountable and effective.
It is written to complement the OWASP Top 10 for Agentic AI (ASI01–ASI10), supplying the behavioral threat layer those technical and architectural categories assume but don't model.
What's inside. The shift governance hasn't caught up to, the behavioral blind spot in agentic AI discourse, the five mechanisms, a way to make them measurable, organizational implications, sector-specific manifestations and an appendix with a Behavioral Agentic Threat Model.
Who it's for. AI risk, security and governance leaders, and organizations deploying agents.