An invitation-only conference on governance and worker voice inside frontier AI labs.
Hosted by Baroness Beeban Kidron OBE, Member of the House of Lords. Convened by the Institute for Ethics in AI at the University of Oxford, through its Accelerator Fellowship Programme, and the Center for Labor and a Just Economy at Harvard Law School. Coordinated by Democratic Intelligence.
About the day
Mainstream Artificial Intelligence safety treats “alignment” primarily as an engineering puzzle. In practice, making AI accountable to human values and priorities (i.e. alignment) is also an institutional challenge: frontier systems mirror the priorities of the concentrated capital that finances and directs them. As this race evolves, corporate leadership routinely sidelines and devalues workers’ knowledge, ignoring the very expertise required to steer these systems safely.
A growing cycle of open letters, whistleblower warnings, and high-profile resignations across frontier AI companies reveals that a potentially powerful and credible demand for accountability is originating from within the AI companies themselves. Voluntary corporate pledges, self-selected boards, and discretionary executive promises have proven structurally incapable of checking competitive and massive financial pressures and lack credibility among many policymakers and stakeholders.
We cannot afford to accept the continued expansion of AI without ensuring that those who labour to develop it have a real say, and that their knowledge is empowered to steer both ethical development and safety. In particular, the fact that dominant frontier AI labs appear to have a conflict of interest with regard to regulation is proving to be a significant obstacle to creating a consensus around what regulation is needed and how urgently it is needed. To counter this, policy must establish viable channels for the diverse internal knowledge of employees to actively shape decisions, ensuring that critical worker voice is not reduced to resignation, or “exit” per Albert O. Hirschman’s typology. By anchoring workers’ voice directly into corporate internal governance structures, technical and operational staff will secure the structural leverage needed to audit, steer, and veto frontier models in the public interest. The promise of alignment must be transformed from a discretionary executive pledge into an enforceable standard for AI companies. This structural shift is crucial, and forms a cornerstone of the wider far-reaching regulatory transformation required to ensure technology serves the common good.
Similar empowerment will have to be extended beyond frontier AI companies to cover the whole of society. At work, the same structural guarantees will have to be granted to all those subjected to AI systems across workplaces. As the ones who invest their time, brain, and mental and physical health to solve the firm’s problems, workers are entitled to full algorithmic transparency and accountability; systems cannot answer solely to capital investors while governing the day-to-day lives of the workforce. In society at large, seeking to ensure the people consent to the rules that govern them, appropriate accountability mechanisms to ensure AI is deployed in the service of humanity are urgently needed (see e.g. Tang and Landemore 2026). This represents the next foundational chapter of the social contract in the algorithmic age: people must hold the enforceable democratic right to shape and consent to AI deployment, directly and through their representatives, with the statutory power to validate, reshape, or veto the systems that govern their life, at work and beyond.
This parliamentary day convenes policymakers, frontier engineers, organized labour, business leaders, and legal scholars to deliberate actionable policy pathways.
Four essential strategic angles
To build a coherent regulatory architecture, the agenda tackles AI governance across four interdependent fronts, via the following four-session programme:
- Employee voice and internal whistleblowing: moving beyond forced resignations and non-disclosure gag orders to institutionalized voice, worker protections, and collective agency.
- Technical alignment and safety frontiers: demystifying what alignment actually means in practice, examining upcoming technical safety bottlenecks, and exploring the potential of AI constitutions as a tool for oversight and alignment.
- Corporate governance and structural reform: interrogating whether we can remain satisfied with the corporate status quo, evaluating the limits of elite trusts, and codifying worker voice at board level, and below.
- International and multilateral coordination: formulating joint strategies that national governments and international bodies (such as the UN, ILO and EU) can pursue to prevent a regulatory race to the bottom.
How the day works
Each session is a moderated panel followed by an open floor. Speakers receive the reference points for their session in advance, and the chair summarises them at the start, so the discussion begins from the proposals rather than from first principles.
The opening and the closing pledge session are on the record. The four working sessions run under the Chatham House Rules: what is said may be used, but not attributed.
Two outputs follow. Participants receive a two-page readout within a week, under the agreed basis. The convenors publish a short public memo on lessons learned and next steps from this day in Westminster.
Visit the conference page to view the full programme.
For the full list of speakers, view the conference page.