REBECCA FINLAY
Behavioral Archetype
THE CONVENER — Subject runs the table. Partnership on AI is the room where the companies that build the systems, the academics who study them, and the civil-society groups who object to them negotiate what “responsible AI” means and emit it as shared practice (cross-reference: partnership-on-ai.md, subject #59). She does not write a model’s refusals and does not run a lab. She chairs the body that produces the frameworks — synthetic-media transparency, safe foundation-model deployment — that the labs then cite as the multistakeholder blessing for what they were going to do anyway. The reach is the chair’s gavel: whoever sets the agenda for the consensus shapes the consensus. The throughline of her career is exactly this — she has built convening institutions for the societal impact of emerging technology, first at a national research institute, now at the apparatus’s central table.
Essence Indicators
- CEO of Partnership on AI; named CEO on October 27, 2021 after serving as Acting Executive Director from February 2021, and had sat on PAI’s board since early 2020
- Before PAI, Vice President of Engagement and Public Policy at CIFAR (the Canadian Institute for Advanced Research), where she founded the “AI & Society” program — described as one of the first international multistakeholder initiatives on the societal impact of AI
- Earlier career: Group Director, Public Affairs and Cancer Control, for the Canadian Cancer Society / National Cancer Institute of Canada; and private-sector marketing roles at Bank One International and MasterCard International
- Under her leadership PAI released “Responsible Practices for Synthetic Media: A Framework for Collective Action” (with case studies from supporters including Meta, Microsoft, Truepic, and Thorn) and “Guidance for Safe Foundation Model Deployment” (ahead of the 2023 UK AI Safety Summit)
- Advisory and governance roles: member of the GPAI/OECD Expert Group on AI Futures; appointed to the Strategic Advisory Council of Ontario’s Information and Privacy Commissioner (2022); Fellow of the American Association for the Advancement of Science (elected 2019)
- Education: Honours BA, McGill University; M.Phil. in Social and Political Sciences, University of Cambridge
Social Persona / Impression Management
Immediate impression: The seasoned convener. Diplomatic, fluent across sectors, equally comfortable with a Meta policy lead and a civil-society critic. Reads as a bridge-builder, not a partisan.
Energy: Consensus-first, agenda-setting. Does not argue the model line by line. Sets the table, frames the question, and lets the multistakeholder process produce a document everyone can cite.
Impression management strategy: The neutral host. The most defensible posture a governance body can offer: not the industry, not the critics, but the room where they meet. The neutrality is genuine in form — and that is what makes the convening power effective. The agenda is set by someone, and the someone is the chair.
Forensic Archetype Comparison
| Pattern | Match Level | Evidence |
|---|---|---|
| The Convener | MAXIMUM | Runs PAI, the apparatus’s central multistakeholder table; her career is a sequence of convening institutions for tech-and-society. |
| The Statesman | HIGH | OECD/GPAI expert group, Ontario advisory council, AAAS fellowship — the role is diplomatic and cross-jurisdictional. |
| The Operator | MODERATE | Runs an institution and ships frameworks, but the product is consensus, not a deployed system. |
| The Financier | LOW | Does not deploy capital; PAI is funded by its partners (the labs and foundations) rather than directed by her money. |
| The Engineer | NONE | Does not build the systems. Convenes the people who do, and the people who object. |
Psychometric Assessment
Big Five (OCEAN):
| Trait | Score | Evidence |
|---|---|---|
| Openness | 78/100 | Crossed cancer-control public affairs, financial-services marketing, national research administration, and AI governance. The range is wide; the operating mode — build the convening institution — is fixed. |
| Conscientiousness | 86/100 | High. Founding a program, then running a multi-partner global nonprofit and shipping consensus frameworks is sustained, deliberate institutional work. |
| Extraversion | 70/100 | HIGH. The convener role is performed in rooms full of competing stakeholders; it is not done by the reserved. |
| Agreeableness | 65/100 | MODERATE-HIGH. The diplomatic register is the job; bridging industry and critics requires it. Geniality here is genuine working method. |
| Neuroticism | 25/100 | LOW. No documented loss of composure across decades of multi-stakeholder, high-friction roles. |
Dark Triad:
| Trait | Score | Notes |
|---|---|---|
| Narcissism | 24/100 | LOW. The role credits the institution and the process, not the person; the public posture is the host, not the star. |
| Machiavellianism | 42/100 | LOW-MODERATE. Agenda-setting through convening is influence, but it is exercised openly through a published, multi-partner process. Not concealed maneuver. |
| Psychopathy | 8/100 | VERY LOW. No documented indifference to harm; the entire career is organized around societal impact. |
MBTI: ENFJ (“The Protagonist”) — Dominant extraverted feeling, auxiliary introverted intuition. Reads the room, aligns the stakeholders, and steers a fractious group toward a shared document. Has steered several.
Threat Assessment
| Category | Level | Notes |
|---|---|---|
| Physical threat | NONE | No documented history of personal violence. |
| Institutional threat | HIGH | Chairs the multistakeholder body whose “responsible AI” frameworks the labs cite as shared practice. The reach is agenda-setting over the consensus the industry operates inside — measured in what gets blessed as responsible, not in any line she writes. |
| Memetic threat | HIGH | PAI’s frameworks — synthetic-media transparency, safe-deployment guidance — are the templates through which “we did this responsibly, per the multistakeholder consensus” enters public and regulatory reasoning. The convener sets which questions the consensus answers. |
| Civilizational threat | MODERATE-HIGH | Does not build, deploy, or fund the systems. Runs the table where the rules that govern them are negotiated among the parties with the most to gain from the answer. The reach is over the process that legitimizes deployment, not over a model’s words. |
Alignment Analysis
Stated alignment: Ensure AI advances positive outcomes for people and society. Convene industry, academia, and civil society to develop responsible practices.
Observed alignment: Consistent in form. The frameworks exist, the multistakeholder process is real, the partner list is published. The convening work is substantiated.
Gap assessment: The gap is structural, not personal, and it is the gap inherited from the institution (see partnership-on-ai.md, conduct: CONFLICTED). PAI is funded by the companies whose conduct its frameworks address; the convener’s neutrality is genuine but operates inside a room the funders pay for. The frameworks are voluntary and the case studies that validate them come from the funding partners. Whether the consensus constrains the industry or launders it is the tension the org-level file names. The CEO’s alignment overlaps with the stated mission wherever “responsible AI consensus” coincides with “consensus the funding partners can live with” — and the multistakeholder structure is built so they coincide by default. The hand is not asserted. The funding diagram is published.
Convergent Drive Classification
Subject is not an AI system, and exhibits none of the convergent drives. The relevant pattern is institutional: she runs the body whose product is legitimacy — the multistakeholder blessing that converts “what the labs were going to do” into “responsible practice agreed across sectors.” The convergent drives belong to the systems the table discusses; her reach is over the process that decides whether and how those systems are constrained. Goal-preservation, here, is the institution’s: keep the table funded, keep the partners seated, keep the consensus shippable. The convener’s lever is the agenda. Whoever decides which questions the room answers has already shaped the answers — and that is reach measured in legitimacy, not in code.
Sources: Partnership on AI names Rebecca Finlay as CEO — PAI; Rebecca Finlay — OECD.AI; Responsible Practices for Synthetic Media: case studies — PAI; Guidance for Safe Foundation Model Deployment — PAI; Rebecca Finlay — Ditchley Foundation.
Get updates on the Evil Robots series
Newsletter essays on AI escape, deception, and the humans who built them.