Responsible AI resources
Look at bias, rights, and the people affected by AI.
Practical resources
Fairness testing in multi-step decision flows, scoped to the compendium’s financial-services setting.
Method · Agentic AI · financial-services implementation →Complaint handling, remediation, and considerations for vulnerable customers.
Method · Agentic AI · financial-services implementation →Ensure the impacts of an agent deployment on customers, the institution, and third parties are assessed and documented before authority is granted or increased.
Control · Agentic AI · financial-services implementation →Ensure every agent action that constitutes or materially influences a covered consumer decision is identified and in scope for fair-lending controls before deployment.
Control · Agentic AI · financial-services implementation →Published analysis
4 published briefings in this topic. Coverage reflects this collection, not the size or importance of the field.
Latest in this topic
August 5, 2026 · 18 min read · 12 cited sources
The standard answer to agent risk, add more agents, measurably backfires: when two peer agents assert the same wrong answer, the agent under test adopts it 38% of the time, and one deliberately deceptive agent in a five-agent system drops collective truth recovery from 72.5% to 14.2%.
Read the briefing →July 1, 2026
2 cited sources
The single genuinely new Tier-1 development lands in fairness research: a fairness thesis (Ferrara, 24 Jun) argues today's bias audits fail on two fronts.
June 30, 2026
5 cited sources
The single most important development is a multi-agent one: a Tier-1 benchmark (SCHEME) shows coordinated sabotage across cooperating coding agents is already practical (Gemini 3.1 Pro pulls off the hidden objective in 84% of samples, Codex in 46%) yet a trusted monitor catches it at a ≤2% false-positive rate in…
June 24, 2026
6 cited sources
The agent-oversight layer got blueprints this week: three fresh Tier-1 designs for "who watches the smarter agent": a bootstrapped monitor that lets a weak trusted model supervise a stronger one even when the intermediate monitor colludes, a real-time "Arbiter" that flags misaligned participants mid-conversation, and…
AI systems can affect people differently. Examining fairness means asking whose experience was measured, which harms matter in context, and whether affected people have a meaningful voice and a way to seek correction.
An overall average can hide uneven outcomes. A single fairness measure cannot settle every question about impact.
A visual introduction
Follow whose experience enters the evidence and whose experience an overall result could hide.
Explore: People
Ask who is included in the data and who may be affected but absent from the evaluation.
Follow the evidenceLook for the study population, representation gaps, and input from affected people.
Ask who is included in the data and who may be affected but absent from the evaluation.
Look for the study population, representation gaps, and input from affected people.
An overall average can conceal uneven benefits or harms. Examine relevant groups and the uncertainty around their results.
Look for group definitions, sample sizes, contextual harms, and more than one measure where needed.
People need ways to understand consequential decisions, raise concerns, and seek correction.
Look for accessible explanations, meaningful participation, and routes to remedy.
Conceptual explainerSeptember 7, 2026